IJACSA Volume 14 Issue 6 - thesai.org

IJACSA Volume 14 Issue 6

Copyright Statement: This is an open access publication licensed under a Creative Commons Attribution 4.0 International License, which permits unrestricted use, distribution, and reproduction in any medium, even commercially as long as the original work is properly cited.

View Full Issue

Paper 1: A Fuzzy Reward and Punishment Scheme for Vehicular Ad Hoc Networks

Abstract: Trust management is an important security approach for the successful implementation of Vehicular Ad Hoc Networks (VANETs). Trust models evaluate messages to assign reward or punishment. This can be used to influence a driver’s future behaviour. In the author’s previous work, a sender-side based trust management framework is developed which avoids the receiver evaluation of messages. However, this does not guarantee that a trusted driver will not lie. These “untrue attacks” are resolved by the RSUs using collaboration to rule on a dispute, providing a fixed amount of reward and punishment. The lack of sophistication is addressed in this paper with a novel fuzzy RSU controller considering the severity of incident, driver past behaviour, and RSU confidence to determine the reward or punishment for the conflicted drivers. Although any driver can lie in any situation, it is expected that trustworthy drivers are more likely to remain so, and vice versa. This behaviour is captured in a Markov chain model for sender and reporter drivers where their lying characteristics depend on trust score and trust state. Each trust state defines the driver’s likelihood of lying using different probability distribution. An extensive simulation is performed to evaluate the performance of the fuzzy assessment and examine the Markov chain driver behaviour model with changing the initial trust score of all or some drivers in Veins simulator. The fuzzy and the fixed RSU assessment schemes are compared, and the result shows that the fuzzy scheme can encourage drivers to improve their behaviour.

Author 1: Rezvi Shahariar

Author 2: Chris Phillips

Keywords: VANET; Trust management; fuzzy logic; Markov chain; reward and punishment; driver behaviour model

Paper 2: Investigating the User Experience and Evaluating Usability Issues in AI-Enabled Learning Mobile Apps: An Analysis of User Reviews

Abstract: Integrating artificial intelligence (AI) has become crucial in modern mobile application development. However, the current integration of AI in mobile learning applications presents several challenges regarding mobile app usability. This study aims to identify critical usability issues of AI-enabled mobile learning apps by analyzing user reviews. We conducted a qualitative and content analysis of user reviews for two groups of AI apps from the education category - language learning apps and educational support apps. Our findings reveal that while users generally report positive experiences, several AI-related usability issues impact user satisfaction, effectiveness, and efficiency. These challenges include AI-related functionality issues, performance, bias, explanation, and ineffective Features. To enhance user experience and learning outcomes, developers must improve AI technology and adapt learning methodologies to meet users’ diverse demands and preferences while addressing these issues. By overcoming these challenges, AI-powered mobile learning apps can continue to evolve and provide users with engaging and personalized learning experiences.

Author 1: Bassam Alsanousi

Author 2: Abdulmohsen S. Albesher

Author 3: Hyunsook Do

Author 4: Stephanie Ludi

Keywords: Human-Computer Interaction (HCI); Artificial Intelligence (AI); user reviews; AI-Enabled Mobile Apps; usability

Paper 3: A Hybrid Method Based on Gravitational Search and Genetic Algorithms for Task Scheduling in Cloud Computing

Abstract: Cloud computing has emerged as a novel technology that offers convenient and cost-effective access to a scalable pool of computing resources over the internet. Task scheduling plays a crucial role in optimizing the functionality of cloud services. However, inefficient scheduling practices can result in resource wastage or a decline in service quality due to under- or overloaded resources. To address this challenge, this research paper introduces a hybrid approach that combines gravitational search and genetic algorithms to tackle the task scheduling problem in cloud computing environments. The proposed method leverages the strengths of both gravitational search and genetic algorithms to achieve enhanced scheduling performance. By integrating the unique search capabilities of the gravitational search algorithm with the optimization and adaptation capabilities of the genetic algorithm, a more effective and efficient solution is achieved. The experimental results validate the superiority of the proposed method over existing approaches in terms of total cost optimization. The experimental evaluation demonstrates that the hybrid method outperforms previous scheduling methods in achieving optimal resource allocation and minimizing costs. The improved performance is attributed to the combined strengths of the gravitational search and genetic algorithms in effectively exploring and exploiting the solution space. These findings underscore the potential of the proposed hybrid method as a valuable tool for addressing the task scheduling problem in cloud computing, ultimately leading to improved resource utilization and enhanced service quality.

Author 1: Xiuyan ZHANG

Keywords: Cloud computing; task scheduling; genetic algorithm; gravitational search algorithm

Paper 4: Shape Control of a Dual-Segment Soft Robot using Depth Vision

Abstract: Pneumatic soft robots outperform rigid robots in complex environments due to the high flexibility of their redundant configurations, and their shape control is considered a prerequisite for applications in unstructured environments. In this paper, we propose a depth vision-based shape control method for a two-segment soft robot, which uses a binocular camera to achieve 3D shape control of the soft robot. A closed-loop control algorithm based on depth vision is designed for shape compensation when subject to its own non-linear responsiveness and coupling by solving for the shape feature parameters used to describe the robot and analytically modeling the motion of curved feature points. Experimental results show that the position and angle errors are less than 2 mm and 1° respectively, the curvature error is less than 0.0001mm-1, and the algorithm has convergence performance for L-type and S-type shape reference 3D shapes. This work provides a general method for being able to adjust the shape of a soft robot without on-board sensors.

Author 1: Hu Junfeng

Author 2: Zhang Jun

Keywords: Pneumatic soft robot; shape control; depth vision; shape feature

Paper 5: Fast Pasture Classification Method using Ground-based Camera and the Modified Green Red Vegetation Index (MGRVI)

Abstract: The assessment of aboveground biomass is important for achieving rational usage of pasture resources and for maximizing the quantity and quality of milk and meat production. This study presents a method for fast approximation of pastures’ biomass. Unlike most similar studies, which rely on unmanned aerial vehicle and satellite obtained data, this study focuses on photos made by stationary or mobile ground-based visual spectrum camera. The developed methodology uses raster analysis, based on the MGRVI index, in order to classify the pasture into two categories: “grazed” and “ungrazed”. Thereafter, the developed methodology accounts for the perspective in order to obtain the actual area of each class in square meters and in percent. The methodology was applied on an experimental pasture, located near the city of Troyan (Bulgaria). Two images were selected, with the first one representing a mostly ungrazed pasture and the second one – a mostly grazed one. Thereafter the images were analyzed using QGIS 3.0 as well as a specially developed software tool. An important advantage of the proposed methodology is that it does not require expensive equipment and technological knowledge, as it relies on commonly available tools, such as the camera of mobile phones.

Author 1: Boris Evstatiev

Author 2: Tsvetelina Mladenova

Author 3: Nikolay Valov

Author 4: Tsenka Zhelyazkova

Author 5: Mariya Gerdzhikova

Author 6: Mima Todorova

Author 7: Neli Grozeva

Author 8: Atanas Sevov

Author 9: Georgi Stanchev

Keywords: Pasture biomass; MGRVI; ground-based camera; classification

Paper 6: Instructional Digital Model to Promote Virtual Teaching and Learning for Autism Care Centres

Abstract: The COVID-19 pandemic has led to temporary school closures affecting over 90% of students worldwide. This has exacerbated educational inequality, particularly for students with learning disabilities such as autism spectrum disorder (ASD), whose routines, services, and support they rely on have been disrupted. To address this issue, it is important to investigate virtual teaching and learning (VTL) strategies that can provide a more effective learning experience for these unique learners. The main objectives of this research are twofold: to investigate the challenges faced by teachers and students with ASD in Malaysia when adapting to online education, and to explore how the learning process occurs during the pandemic. Additionally, the study aimed to identify suitable VTL technology for autism care centres. Four autism care centres were visited for on-site observation activities, and interviews were conducted with the care centre principals. Two sets of online questionnaires were distributed to 10 autism care centres, where 6 principals and 16 teachers provided feedback. The data collected through on-site observations, interviews, and online questionnaires, were then analysed to construct an instructional digital model (IDM) for VTL. The model is very significant as a guide for the development of VTL platform for autism care centres. Finally, a VTL platform development framework was created, which provides a structure for system developers to conduct further research on the development of VTL platform based on the IDM. The framework aims to facilitate the successful implementation of the VTL.

Author 1: Norziana Yahya

Author 2: Nazean Jomhari

Author 3: Mohd Azahani Md Taib

Author 4: Nahdatul Akma Ahmad

Keywords: Instructional digital model; virtual teaching and learning; autism; online learning; pandemic

Paper 7: Investigating OpenAI’s ChatGPT Potentials in Generating Chatbot's Dialogue for English as a Foreign Language Learning

Abstract: Lack of opportunities is a significant hurdle for English as a Foreign Language (EFL) for students during their learning journey. Previous studies have explored the use of chatbots as learning partners to address this issue. However, the success of chatbot implementation depends on the quality of the reference dialogue content, yet research focusing on this subject is still limited. Typically, human experts are involved in creating suitable dialogue materials for students to ensure the quality of such content. Research attempting to utilize artificial intelligence (AI) technologies for generating dialogue practice materials is relatively limited, given the constraints of existing AI systems that may produce incoherent output. This research investigates the potential of leveraging OpenAI's ChatGPT, an AI system known for producing coherent output, to generate reference dialogues for an EFL chatbot system. The study aims to assess the effectiveness of ChatGPT in generating high-quality dialogue materials suitable for EFL students. By employing multiple readability metrics, we analyze the suitability of ChatGPT-generated dialogue materials and determine the target audience that can benefit the most. Our findings indicate that ChatGPT's dialogues are well-suited for students at the Common European Framework of Reference for Languages (CEFR) level A2 (elementary level). These dialogues are easily comprehensible, enabling students at this level to grasp most of the vocabulary used. Furthermore, a substantial portion of the dialogues intended for CEFR B1 (intermediate level) provides ample stimulation for learning new words. The integration of AI-powered chatbots in EFL education shows promise in overcoming limitations and providing valuable learning resources to students.

Author 1: Julio Christian Young

Author 2: Makoto Shishido

Keywords: ChatGPT; chatbots as learning partners; EFL chatbot system; dialogue creation

Paper 8: ConvNeXt-based Mango Leaf Disease Detection: Differentiating Pathogens and Pests for Improved Accuracy

Abstract: Mango farming is a key economic activity in several locations across the world. Mango trees are prone to various diseases caused by viruses and pests, which can substantially impair crops and have an effect on farmers' revenue. To stop the spread of these illnesses and to lessen the crop damage they cause, early diagnosis of these diseases is essential. Growing interest has been shown in employing deep learning models to create automated disease detection systems for crops because of recent developments in machine learning. This research article includes a study on the application of ConvNeXt models for the diagnosis of pathogen and pest caused illnesses in mango plants. The study intends to investigate the variety in how these illnesses emerge on mango leaves and assess the efficiency of ConvNeXt models in identifying and categorizing them. Images of healthy mango leaves as well as the leaves with a variety of illnesses brought on by pathogens and pests are included in the dataset used in the study. In the study, deep learning models were applied to classify mango pests and pathogens. The models achieved high accuracy on both datasets, with better performance on the pathogen dataset. Larger models consistently outperformed smaller ones, indicating their ability to learn complex features. The ConvNeXtXLarge model showed the highest accuracy: 98.79% for mango pests, 100% for mango pathogens, and 99.17% for the combined dataset. This work holds significance for mango disease detection, aiding in efficient management and potential economic benefits for farmers. However, the models' performance can be influenced by dataset quality, preprocessing techniques, and hyperparameter selection.

Author 1: Asha Rani K P

Author 2: Gowrishankar S

Keywords: Mango disease; pest; pathogens; machine learning; deep learning; convnext models

Paper 9: A Fine-grained Access Control Model with Enhanced Flexibility and On-chain Policy Execution for IoT Systems

Abstract: Blockchain-based access control mechanisms have garnered significant attention in recent years due to their potential to address the security and privacy challenges in the Internet of Things (IoT) ecosystem. IoT devices generate massive amounts of data that are often transmitted to cloud-based servers for processing and storage. However, these devices are vulnerable to attacks and unauthorized access, which can lead to data breaches and privacy violations. Blockchain-based access control mechanisms can provide a secure and decentralized solution to these issues. This paper presents an improved Attribute-based Access Control (ABAC) approach with enhanced flexibility, which utilizes decentralized identity management on the Substrate Framework, codifies access control policies by Rust programming language, and executes access control policies on-chain. The proposed design ensures trust and security while enhancing flexibility compared to existing works. In addition, we implement a PoC to demonstrate the feasibility and investigate its effectiveness.

Author 1: Hoang-Anh Pham

Author 2: Ngoc Nhuan Do

Author 3: Nguyen Huynh-Tuong

Keywords: Attribute-based Access Control (ABAC); Internet of Things (IoT); blockchain; substrate framework

Paper 10: DeLClustE: Protecting Users from Credit-Card Fraud Transaction via the Deep-Learning Cluster Ensemble

Abstract: Fraud is the unlawful acquisition of valuable assets gained via intended misrepresentation. It is a crime committed by either an internal/external user, and associated with acts of theft, embezzlement, and larceny. The proliferation of credit cards to aid financial inclusiveness has its usefulness alongside it attracting malicious attacks for gains. Attempts to classify fraudulent credit card transactions have yielded formal taxonomies as these attacks seek to evade detection. We propose a deep learning ensemble via a profile hidden Markov model with a deep neural network, which is poised to effectively classify credit-card fraud with a high degree of accuracy, reduce errors, and timely fashion. The result shows the ensemble effectively classified benign transactions with a precision of 97 percent. Thus, we posit a new scheme that is more logical, intuitive, reusable, exhaustive, and robust in classifying such fraudulent transactions based on the attack source, cause(s), and attack time gap.

Author 1: Fidelis Obukohwo Aghware

Author 2: Rume Elizabeth Yoro

Author 3: Patrick Ogholoruwami Ejeh

Author 4: Christopher Chukwufunaya Odiakaose

Author 5: Frances Uche Emordi

Author 6: Arnold Adimabua Ojugo

Keywords: Fraud transactions; fraud detection; deep learning ensemble; credit card fraud; cluster modeling; financial inclusion

Paper 11: Proof of Spacetime as a Defensive Technique Against Model Extraction Attacks

Abstract: When providing a service that utilizes a machine learning model, the countermeasures against cyber-attacks are required. The model extraction attack is one of the attacks, in which an attacker attempts to replicate the model by obtaining a large number of input-output pairs. While a defense using Proof of Work has already been proposed, an attacker can still conduct model extraction attacks by increasing their computational power. Moreover, this approach leads to unnecessary energy consumption and might not be environmentally friendly. In this paper, the defense method using Proof of Spacetime instead of Proof of Work is proposed to reduce the energy consumption. The Proof of Spacetime is a method to impose spatial and temporal costs on the users of the service. While the Proof of Work makes a user to calculate until permission is granted, the Proof of Spacetime makes a user to keep a result of calculation, so the energy consumption is reduced. Through computer simulations, it was found that systems with Proof of Spacetime, compared to those with Proof of Work, impose 0.79 times the power consumption and 1.07 times the temporal cost on the attackers, while 0.73 times and 0.64 times on the non-attackers. Therefore, the system with Proof of Spacetime can prevent model extraction attacks with lower energy consumption.

Author 1: Tatsuki Fukuda

Keywords: Proof of spacetime; model extraction attacks; machine learning; security

Paper 12: Deep Learning-based Intrusion Detection: A Novel Approach for Identifying Brute-Force Attacks on FTP and SSH Protocol

Abstract: As networks continue to expand rapidly, the number and diversity of cyberattacks are also increasing, posing a significant challenge for organizations worldwide. Consequently, brute-force attacks targeting FTP and SSH protocols have become more prevalent. IDSes offer an essential tool to detect these attacks, providing traffic analysis and system monitoring. Traditional IDSes employ signatures and anomalies to monitor information flow for malicious activity and policy violations; however, they often struggle to effectively identify unknown or novel patterns. In response, we propose a novel intelligent approach based on deep learning to detect brute-force attacks on FTP and SSH protocols. We conducted an extensive literature review and developed a metric to compare our work with existing literature. Our findings indicate that our proposed approach achieves an accuracy of 99.9%, outperforming other comparable solutions in detecting brute-force attacks.

Author 1: Noura Alotibi

Author 2: Majid Alshammari

Keywords: Artificial neural networks; machine learning; deep ‎learning; intrusion detection ‎system; detecting brute ‎force attacks on SSH and FTP protocols

Paper 13: Method for Characterization of Customer Churn Based on LightBGM and Experimental Approach for Mitigation of Churn

Abstract: A method for customer churn characterization based on LightBGM (Light Gradient Boosting Machine) is proposed. Additionally, experimental approaches for mitigation of churn are conducted through churn prediction. The experiments reveal several churn characteristics such as age dependency, gender dependency (with a high divorce rate among female customers), number of visits dependency (with a higher churn rate for customers with fewer visits), unit price (per hair salon visit) dependency (with a higher withdrawal rate for lower-priced services), date of first visit dependency (with a high churn rate for recent customers), date of last visit dependency, and menu dependency (with low attrition rates for gray hair dye and high attrition rates for school and child cuts) and so on. Through the experiments, these dependencies are clarified. It is found that the first visit date is the most significant factor for churn customer character. Also, it is found that “distance to hair salon” dependency may be related to the availability of parking lots, although this factor has insignificant impact on the churn rate.

Author 1: Kohei Arai

Author 2: Ikuya Fujikawa

Author 3: Yusuke Nakagawa

Author 4: Ryoya Momozaki

Author 5: Sayuri Ogawa

Keywords: Churn; LightBGM; churn characteristics; linear regression

Paper 14: Artificial Intelligence-based Detection of Fava Bean Rust Disease in Agricultural Settings: An Innovative Approach

Abstract: The traditional methods used to identify plant diseases mostly rely on expert opinion, which causes long waits and enormous expenses in the control of crop diseases and field activities, especially given that the majority of crop infections now in existence have tiny targets, occlusions, and looks that are similar to those of other diseases. To increase the efficiency and precision of rust disease classification in a fava bean field, a new optimized multilayer deep learning model called YOLOv8 is suggested in this study. 3296 images were collected from a farm in eastern Morocco for the fava bean rust disease dataset. We labeled all the data before training, evaluating, and testing our model. The results demonstrate that the model developed using transfer learning has a higher recognition precision than the other models, reaching 95.1%, and can classify and identify diseases into three severity levels: healthy, moderate, and critical. As performance indicators, the needed standards for mean Average Precision (mAP), recall, and F1 score are 93.7%, 90.3%, and 92%, respectively. The improved model's detection speed was 10.1 ms, sufficient for real-time detection. This study is the first to employ a new method to find rust in fava bean crops. Results are encouraging and supply new opportunities for crop disease research.

Author 1: Hicham Slimani

Author 2: Jamal El Mhamdi

Author 3: Abdelilah Jilbab

Keywords: Fava bean disease; deep learning; YOLOv8; real-time detection

Paper 15: Bidirectional Long-Short-Term Memory with Attention Mechanism for Emotion Analysis in Textual Content

Abstract: Emotion analysis in textual content plays a crucial role in various applications, including sentiment analysis, customer feedback monitoring, and mental health assessment. Traditional machine learning and deep learning techniques have been employed to analyze emotions; however, these methods often fail to capture complex and long-range dependencies in text. To overcome these limitations, this paper proposes a novel bidirectional long-short-term memory (Bi-LSTM) model for emotion analysis in textual content. The proposed Bi-LSTM model leverages the power of recurrent neural networks (RNNs) to capture both the past and future context of text, providing a more comprehensive understanding of the emotional content. By integrating the forward and backward LSTM layers, the model effectively learns the semantic representations of words and their dependencies in a sentence. Additionally, we introduce an attention mechanism to weigh the importance of different words in the sentence, further improving the model's interpretability and performance. To evaluate the effectiveness of our Bi-LSTM model, we conduct extensive experiments on Kaggle Emotion detection dataset. The results demonstrate that our proposed model outperforms several state-of-the-art baseline methods, including traditional machine learning algorithms, such as support vector machines and naive Bayes, as well as other deep learning approaches, like CNNs and vanilla LSTMs.

Author 1: Batyrkhan Omarov

Author 2: Zhandos Zhumanov

Keywords: Deep learning; emotion detection; BiLSTM; machine learning; classification; artificial intelligence

Paper 16: Artificial Intelligence Enabled Mobile Chatbot Psychologist using AIML and Cognitive Behavioral Therapy

Abstract: In recent years, the demand for mental health services has increased exponentially, prompting the need for accessible, cost-effective, and efficient solutions. This paper introduces an Artificial Intelligence (AI) enabled mobile chatbot psychologist that leverages AIML (Artificial Intelligence Markup Language) and Cognitive Behavioral Therapy (CBT) to provide psychological support. The chatbot is designed to facilitate mental health care by offering personalized CBT interventions to individuals experiencing psychological distress. The proposed mobile chatbot psychologist employs AIML, a language created to facilitate human-computer interactions, to understand user inputs and generate contextually appropriate responses. To ensure the efficacy of the chatbot, it is equipped with a knowledge base comprising CBT principles and techniques, enabling it to provide targeted psychological interventions. The integration of CBT allows the chatbot to address a wide range of mental health issues, including anxiety, depression, stress, and phobias, by helping users identify and challenge cognitive distortions. The paper discusses the development and implementation of the mobile chatbot psychologist, detailing the AIML-based conversational engine and the incorporation of CBT techniques. The chatbot's effectiveness is evaluated through a series of user studies involving participants with varying levels of psychological distress. Results demonstrate the chatbot's ability to deliver personalized interventions, with users reporting significant improvements in their mental well-being. The AI-enabled mobile chatbot psychologist offers a promising solution to bridge the gap in mental health care, providing an easily accessible, cost-effective, and scalable platform for psychological support. This innovative approach can serve as a valuable adjunct to traditional therapy and help reduce the burden on mental health professionals, while empowering individuals to take charge of their mental well-being.

Author 1: Batyrkhan Omarov

Author 2: Zhandos Zhumanov

Author 3: Aidana Gumar

Author 4: Leilya Kuntunova

Keywords: Chatbot; artificial intelligence; machine learning; CBT; AIML

Paper 17: A Multi-branch Feature Fusion Model Based on Convolutional Neural Network for Hyperspectral Remote Sensing Image Classification

Abstract: Hyperspectral image classification constitutes a pivotal research domain in the realm of remote sensing image processing. In the past few years, convolutional neural networks (CNNs) with advanced feature extraction capabilities have demonstrated remarkable performance in hyperspectral image classification. However, the challenges faced by classification methods are compounded by the difficulties of "dimensional disaster" and limited sample distinctiveness in hyperspectral images. Despite existing efforts to extract spectral spatial information, low classification accuracy remains a persistent issue. Therefore, this paper proposes a multi-branch feature fusion model classification method based on convolutional neural networks to fully extract more effective and adequate high-level semantic features. The proposed classification model first undergoes PCA dimensionality reduction, followed by a multi-branch network composed of three-dimensional and two-dimensional convolutions. Convolutional kernels of varying scales are utilized for multi-feature extraction. Among them, the 3D convolution not only adapts to the cube of hyperspectral data but also fully exploits the spectral-spatial information, while the 2D convolution learns deeper spatial information. The experimental results of the proposed model on three datasets demonstrate its superior performance over traditional classification models, enabling it to accomplish the task of hyperspectral image classification more effectively.

Author 1: Jinli Zhang

Author 2: Ziqiang Chen

Author 3: Yuanfa Ji

Author 4: Xiyan Sun

Author 5: Yang Bai

Keywords: Hyperspectral image classification; convolutional neural network (CNN); multi-branch network; feature fusion

Paper 18: Socio Technical Framework to Improve Work Behavior During Smart City Implementation

Abstract: Every organization uniquely adheres to security culture. Numerous studies have discovered that procrastinating, impulsive, forward-thinking, and risk-taking behaviors vary across organizations, which may help to explain why different organizations' adherence to security policies. This study describes the human aspect of a government organization in contributing to the successful implementation of a smart city by minimizing cybersecurity threats. Improper employee behavior and lack of understanding of cybersecurity will negatively contribute to the successful development of smart cities. The purpose of this research is to develop a framework to determine the factors to improve work behavior in terms of the contribution of social and technical factors. The use of a socio-technical approach to explain how socio-technical integration can contribute to improving work behavior by using mixed methods. The results indicated that several socio-technical factors which include technology, IT infrastructure, work organization, competency, training, and teamwork contribute to improving work behaviors which can be used as a basis for minimizing cybersecurity threats in smart city implementation.

Author 1: Eko Haryadi

Author 2: Abdul Karim

Author 3: Lizawati Salahuddin

Keywords: Framework; socio technical; cybersecurity; behavior; threat; smart city

Paper 19: Detecting Malware with Classification Machine Learning Techniques

Abstract: In today's digital landscape, the identification of malicious software has become a crucial undertaking. The ever-growing volume of malware threats renders conventional signature-based methods insufficient in shielding against novel and intricate attacks. Consequently, machine learning strategies have surfaced as a viable means of detecting malware. The following research report focuses on the implementation of classification machine learning methods for detecting malware. The study assesses the effectiveness of several algorithms, including Naïve Bayes, Support Vector Machine (SVM), K-Nearest Neighbor (KNN), Decision Tree, Random Forest, and Logistic Regression, through an examination of a publicly accessible dataset featuring both benign files and malware. Additionally, the influence of diverse feature sets and preprocessing techniques on the classifiers' performance is explored. The outcomes of the investigation exhibit that machine learning methods can capably identify malware, attaining elevated precision levels and decreasing false positive rates. Decision Tree and Random Forest display superior performance compared to other algorithms with 100.00% accuracy. Furthermore, it is observed that feature selection and dimensionality reduction techniques can notably enhance classifier effectiveness while mitigating computational complexity. Overall, this research underscores the potential of machine learning approaches for detecting malware and offers valuable guidance for the development of successful malware detection systems.

Author 1: Mohd Azahari Mohd Yusof

Author 2: Zubaile Abdullah

Author 3: Firkhan Ali Hamid Ali

Author 4: Khairul Amin Mohamad Sukri

Author 5: Hanizan Shaker Hussain

Keywords: Malware; classification; machine learning; accuracy; false positive rate

Paper 20: Evaluation of the Accidents Risk Caused by Truck Drivers using a Fuzzy Bayesian Approach

Abstract: Road accidents cause hundreds of fatalities and injuries each year; due to their size and operating features, heavy trucks typically experience more severe accidents. Many factors are likely to cause such accidents; however, statistics mainly blame human error. This paper analyses the risk of accidents for heavy vehicles, focusing on driver-related factors contributing to accidents. A model is developed to anticipate the probability of an accident by using Bayesian networks (BNs) and fuzzy logic. Three axioms were verified to validate the developed model, and a sensitivity analysis is performed to identify the factors that have the most significant influence over truck accidents. Subsequently, the result provided by the model was exploited to examine the effects of in-vehicle road safety systems in preventing road accidents via an event tree analysis. The results underlined a strong link between the occurrence of accidents and parameters related to the driver, such as alcohol and substance consumption, his driving style, and his reactivity. Similarly, unfavourable working conditions significantly impact the occurrence of accidents since it contributes to fatigue, one of the leading causes of road accidents. Also, the event tree analysis results have highlighted the importance of equipping trucks with these mechanisms.

Author 1: Imane Benallou

Author 2: Abdellah Azmani

Author 3: Monir Azmani

Keywords: Heavy truck vehicle; road accident prevention; risk management; bayesian-fuzzy network; analysis tree event

Paper 21: Software Cost Estimation using Stacked Ensemble Classifier and Feature Selection

Abstract: Predicting the cost of the development effort is essential for successful projects. This helps software project managers to allocate resources, and determine budget or delivery date. This paper evaluates a set of machine learning algorithms and techniques in predicting the development cost of software projects. A feature selection algorithm is utilized to enhance the accuracy of the prediction process. A set of evaluations are presented based on basic classifiers and stacked ensemble classifiers with and without the feature selection approach. The evaluation study uses a dataset from 76 university students' software projects. Results show that using a stacked ensemble classifier and feature selection technique can increase the accuracy of software cost prediction models.

Author 1: Mustafa Hammad

Keywords: Software project management; effort estimation; prediction model; machine learning

Paper 22: An Algorithm Based on Self-balancing Binary Search Tree to Generate Balanced, Intra-homogeneous and Inter-homogeneous Learning Groups

Abstract: This paper presents an algorithm, based on the self-balancing binary search tree, to form learning groups. It aims to generate learning groups that are intra-homogeneous (student performance similarity within the group), inter-homogeneous (group performance similarity between groups), and of balanced size. The algorithm mainly uses the 2-3 tree and the 2-3-4 tree as two implementations of a self-balancing binary search tree to form student blocks with close GPAs (grade point averages) and balanced sizes. Then, groups are formed from those blocks in a greedy manner. The experiment showed the efficiency of the proposed algorithm, compared to traditional forming methods, in balancing the size of the groups and improving their intra- and inter-homogeneity by up to 26%, regardless of the used version of the self-balancing binary search tree (2-3 or 2-3-4). For small samples of students, the use of the 2-3-4 tree was distinguished for improving intra- and inter-homogeneity compared to the 2-3 tree. As for large samples of students, experiments showed that the 2-3 tree was better than the 2-3-4 tree in improving the inter-homogeneity, while the 2-3-4 tree was distinguished in improving the intra-homogeneity.

Author 1: Ali Ben Ammar

Author 2: Amir Abdalla Minalla

Keywords: Learning group formation; balanced size groups; homogeneous groups; self-balancing binary search trees; greedy algorithm

Paper 23: Multi-Features Audio Extraction for Speech Emotion Recognition Based on Deep Learning

Abstract: The increasing need for human interaction with computers makes the interaction process more advanced, one of which is by utilizing voice recognition. Developing a voice command system also needs to consider the user's emotional state because the users indirectly treat computers like humans in general. By knowing the type of a person's emotions, the computer can adjust the type of feedback that will be given so that the human-computer interaction (HCI) process will run more humanely. Based on the results of previous research, increasing the accuracy of recognizing the types of human emotions is still a challenge for researchers. This is because not all types of emotions can be expressed equally, especially differences in language and cultural accents. In this study, it is proposed to recognize speech-based emotion types using multi-feature extraction and deep learning. The dataset used is taken from the RAVDESS database. The dataset was then extracted using MFCC, Chroma, Mel-Spectrogram, Contrast, and Tonnetz. Furthermore, in this study, PCA (Principal Component Analysis) and Min-Max Normalization techniques will be applied to determine the impact resulting from the application of these techniques. The data obtained from the pre-processing stage is then used by the Deep Neural Network (DNN) model to identify the types of emotions such as calm, happy, sad, angry, neutral, fearful, surprised, and disgusted. The model testing process uses the confusion matrix technique to determine the performance of the proposed method. The test results for the DNN model obtained the accuracy value of 93.61%, a sensitivity of 73.80%, and a specificity of 96.34%. The use of multi-features in the proposed method can improve the performance of the model's accuracy in determining the type of emotion based on the RAVDESS dataset. In addition, using the PCA method also provides an increase in pattern correlation between features so that the classifier model can show performance improvements, especially accuracy, specificity, and sensitivity.

Author 1: Jutono Gondohanindijo

Author 2: Muljono

Author 3: Edi Noersasongko

Author 4: Pujiono

Author 5: De Rosal Moses Setiadi

Keywords: Deep learning; multi-features extraction; RAVDESS; speech emotion recognition

Paper 24: Hierarchical Convolutional Neural Networks using CCP-3 Block Architecture for Apparel Image Classification

Abstract: In fashion applications, deep learning has been applied automatically to recognize and classify the apparel images under the massive visual data, emerged on social networks. To classify the apparel correctly and quickly is challenging due to a variety of apparel features and complexity of the classification. Recently, the hierarchical convolutional neural networks (H–CNN) with the VGGNet architecture was proposed to classify the fashion-MNIST datasets. However, the VGGNet (many layers) required many filters (in the convolution layer) and many neurons (in the fully connected layer), leading to computational complexity and long training-time. Therefore, this paper proposes to classify the apparel images by the H–CNN in cooperated with the new shallow-layer CCP-3-Block architecture, where each building block consists of two convolutional layers (CC) and one pooling layer (P). In the CCP-3-Block, the number of layers can be reduced (in the network), the number of filters (in the convolution layer), and the number of neurons (in the fully connected layer), while adding a new connection between the convolution layer and the pooling layer plus a batch-normalization technique before passing the activation so that networks can learn independently and train quickly. Moreover, dropout techniques were utilized in the feature mapping and fully connected to reduce overfitting, and the optimizer adaptive moment estimation was utilized to solve the decaying of gradients, which can improve the network-performance. The experimental results showed that the improved H–CNN model with our CCP-3-Block outperformed the recent H–CNN model with the VGGNet in terms of decreased loss, increased accuracy, and faster training.

Author 1: Natthamon Chamnong

Author 2: Jeeraporn Werapun

Author 3: Anantaporn Hanskunatai

Keywords: Convolutional neural networks (CNN); hierarchical CNN (H-CNN); CCP-3 block (two convolutional layers (CC) and one pooling layer (P) per block); apparel image classification; fashion applications

Paper 25: Towards Point Cloud Classification Network Based on Multilayer Feature Fusion and Projected Images

Abstract: Deep Learning (DL) based point cloud classification techniques now in use suffer from issues such as disregarding local feature extraction, missing connections between points, and failure to extract two-dimensional information features from point clouds. A point cloud classification network that utilizes multi-layer feature fusion and point cloud projection images is suggested to address the aforementioned problems and produce more accurate classification outcomes. Firstly, the network extracts local characteristics of point clouds through graph convolution to strengthen the connection between points. Then, the fusing attention mechanism is introduced to aggregate the useful characteristics of the point cloud while suppressing the useless characteristics, and the point cloud characteristics are fused by multi-layer characteristic fusion. Finally, a 3D point cloud network plug-in model based on point cloud projection image (3D CLIP) is proposed, which can make up for the defects of other 3D point cloud classification networks that do not extract two-dimensional information characteristics of point clouds, and solve the problem of low accuracy of similar category recognition in datasets. The ModelNet40 dataset was used for classification studies, and the results show that the point cloud classification network, without the addition of a 3D CLIP plug-in model, achieves a classification accuracy of 92.5%. The point cloud classification network with a 3D CLIP plug-in model achieved a classification accuracy of 93.6%, demonstrating that this technique can successfully raise point cloud classification accuracy.

Author 1: Tengteng Song

Author 2: YiZhi He

Author 3: Muhammad Tahir

Author 4: Jianbo Li

Author 5: Zhao Li

Author 6: Imran Saeed

Keywords: Point cloud; classification; graph convolution; attention mechanism; CLIP

Paper 26: Early Detection of Autism Spectrum Disorder (ASD) using Traditional Machine Learning Models

Abstract: Autism Spectrum Disorder (ASD) is a mental disorder among children that is difficult to diagnose at an early age of a child. People with ASD have difficulty functioning in areas such as communication, social interaction, motor skills, and emotional regulation. They may also have difficulty processing sensory information and have difficulty understanding language, which can lead to further difficulty in socializing. Early detection can help with learning coping skills, communication strategies, and other interventions that can make it easier for them to interact with the world. This kind of disorder is not curable but it is possible to reduce the symptoms of ASD. The early age detection of ASD helps to start several therapies corresponding to ASD symptoms. The detection of ASD symptoms at an early age of a child is our main problem where traditional machine learning algorithms like Support Vector Machine, Logistic Regression, K-nearest neighbour, and Random Forest classifiers have been applied to parents’ dialog to understand the sentiment of each statement about their child. After completion of the prediction of these models, each positive ASD symptoms-related sentence has been used in the cosine similarity model for the detection of ASD problems. Samples of parents’ dialogs have been collected from social networks and special child training institutes. Data has been prepared according to the model for sentiment analysis. The accuracies of these proposed classifiers are 71%, 71%, 62%, and 69% percent according to the prepared data. Another dataset has been prepared where each sentence refers to a particular categorical ASD problem and that has been used in cosine similarity calculation for ASD problem detection.

Author 1: Prasenjit Mukherjee

Author 2: Sourav Sadhukhan

Author 3: Manish Godse

Author 4: Baisakhi Chakraborty

Keywords: Support vector; logistic regression; cosine similarity; K-nearest neighbor; random forest

Paper 27: Speaker Recognition Improvement for Degraded Human Voice using Modified-MFCC with GMM

Abstract: Speaker’s audio is one of the unique identities of the speaker. Nowadays not only humans but machines can also identify humans by their audio. Machines identify different audio properties of the human voice and classify speaker from speaker’s audio. Speaker recognition is still challenging with degraded human voice and limited dataset. Speaker can be identified effectively when feature extraction from voice is more accurate. Mel-Frequency Cepstral Coefficient (MFCC) is mostly used method for human voice feature extraction. We are introducing improved feature extraction method for effective speaker recognition from degraded human audio signal. This article presents experiment results of modified MFCC with Gaussian Mixture Model (GMM) on uniquely developed degraded human voice dataset. MFCC uses human audio signal and transforms it into a numerical value of audio characteristics, which is utilized to recognize speaker efficiently with the help of data science model. Experiment uses degraded human voice when high background noise comes with audio signal. Experiment also covers, Sampling Frequency (SF) impacts on human audio when “Signal to Noise Ratio” (SNR) is low (up to 1dB) in overall speaker identification process. With modified MFCC, we have observed improved speaker recognition when speaker voice SNR is upto 1dB due to high SF and low frequency range for mel-scale triangular filter.

Author 1: Amit Moondra

Author 2: Poonam Chahal

Keywords: GMM; artificial intelligence; MFCC; fundamental frequency; melspectrum; speaker recognition

Paper 28: Application of Conv-1D and Bi-LSTM to Classify and Detect Epilepsy in EEG Data

Abstract: EEG is used to study the electrical changes in the brain and can derive a conclusion as epileptic or not, using an automated method for accurate detection of seizures. Deep learning, a technique ahead of machine learning tools, can self-discover related data for the detection and classification of EEG analysis. Our work focuses on deep neural network architecture to visualize the temporal dependencies in EEG signals. Algorithms and models based on Deep Learning techniques like Conv1D, Conv1D + LSTM, and Conv1D + Bi-LSTM for binary and multiclass classification. Convolution Neural Networks can spontaneously extract and learn features independently in the multichannel time-series EEG signals. Long Short-Term Memory (LSTM) network, with its selective memory retaining capability, Fully Connected (FC) layer, and softmax activation, discover hidden sparse features from EEG signals and predicts labels as output. Two independent LSTM networks combine to form Bi-LSTM in opposite directions and appreciate added visibility to upcoming information to provide efficient work contrary to previous methods. Long-term EEG recordings on the Bonn EEG database, Hauz Khas epileptic database, and Epileptic EEG signals from Spandana Hospital, Bangalore, assess performance. Metrics like precision, recall, f1-score, and support exhibit an improvement over traditional ML algorithms evaluated in the literature.

Author 1: Chetana R

Author 2: A Shubha Rao

Author 3: Mahantesh K

Keywords: 1D CNN; bidirectional LSTM; dataset (DS); deep learning; electroencephalogram (EEG); LSTM

Paper 29: A New Fuzzy Lexicon Expansion and Sentiment Aware Recommendation System in e-Commerce

Abstract: Customers’ feedbacks are necessary for an online business to enrich themselves. The customers’ feedback reflects the quality of the products and the e-commerce services. The companies are in a position to concentrate more and analyze the customers’ feedback or reviews carefully by applying new techniques for predicting the current trends, customers’ expectations, and the quality of their services. The e-business will succeed when one accurately predicts customer purchase patterns and expectations. For this purpose, we propose a new fuzzy logic incorporated sentiment analysis-based product recommendation system to predict the customers’ needs and recommend suitable products successfully. The proposed system incorporates a newly developed sentiment analysis model which incorporates the classification through fuzzy temporal rules. Moreover, the basic level data preprocessing activities such as stemming, stop word removal, syntax analysis and tokenization are performed to enhance the sentiment classification accuracy. Finally, this product recommendation system recommends suitable products to the customers by predicting the customers’ needs and expectations. The proposed system is evaluated using the Amazon dataset and proved better than the existing recommendation systems regarding precision, recall, serendipity and nDCG.

Author 1: Manikandan. B

Author 2: Rama. P

Author 3: Chakaravarthi. S

Keywords: Classification; e-commerce; preprocessing; recommendation system; recurrent neural network; sentiment analysis

Paper 30: Information Technology Technical Support Success Factors in Higher Education: Principal Component Analysis

Abstract: The use of information and communication technologies at higher education institutions is no longer an option, but rather a need. Information Technology support is an essential factor that entails giving end users assistance with hardware and software components. Technical support for information technology has been recognized as a crucial element linked to student happiness because it helps students understand, access, and use technology efficiently. The successful implementation of IT technical support will be aided by identifying the essential success criteria that enable efficient and effective support for students and instructors. Hence the main aim of this study is to identify and rank the key success factors for the successful implementation of IT technical support at higher education institutes. 81 key success factors identified from 100 research papers were analyzed using principal component analysis. The findings led to the identification and ranking of 25 PCs. 95.35 percent of the observed variation was accounted for by the first 25 PCs with eigenvalues higher than 1. The percentages for the first 6 PCs were, in order, 11.87%, 22.21%, 30,64%, 38.25%, 45,12%, and 51.47%. This research provides useful information highlighting factors that can be used to examine areas in educational institutions that need to receive continuous and special care to generate high student satisfaction; ensure future success and gain a competitive advantage. These factors can assist the management of HEI to determine the success or failure of an institution in terms of the technical support provided to students and student satisfaction.

Author 1: Geeta Pursan

Author 2: Timothy. T. Adeliyi

Author 3: Seena Joseph

Keywords: Information technology; technical support services; key success factors; principal component analysis; higher education institutions

Paper 31: Effect of Distance and Direction on Distress Keyword Recognition using Ensembled Bagged Trees with a Ceiling-Mounted Omnidirectional Microphone

Abstract: Audio surveillance can provide an effective alternative to video surveillance in situations where the latter is impractical. Nevertheless, it is essential to note that audio recording raises privacy and legal concerns that require unambiguous consent from all parties involved. By utilizing keyword recognition, audio recordings can be filtered, allowing for the creation of a surveillance system that is activated by distress keywords. This paper investigates the performance of the Ensemble Bagged Trees (EBT) classifier in recognizing the distress keyword "Please" captured by a ceiling-mounted omnidirectional microphone in a room measuring 4.064m (length) x 2.54m (width) x 2.794m (height). The study analyzes the impact of different distances (0m, 1m, and 2m) and two directions (facing towards and away from the microphone) on recognition performance. Results indicate that the system is more sensitive and better able to identify targeted signals when they are farther away and facing toward the microphone. The validation process demonstrates excellent accuracy, precision, and recall values exceeding 98%. In testing, the EBT achieved a satisfactory recall rate of 86.7%, indicating moderate sensitivity, and a precision of 97.7%, implying less susceptibility to false alarms, a crucial feature of any reliable surveillance system. Overall, the findings suggest that a single omnidirectional microphone equipped with an EBT classifier is capable of detecting distress keywords in a low-noise enclosed room measuring up to 4.0 meters in length, 4.0 meters in width, and 2.794 meters in height. This study highlights the potential of employing an omnidirectional microphone and EBT classifier as an edge audio surveillance system for indoor environments.

Author 1: Nadhirah Johari

Author 2: Mazlina Mamat

Author 3: Yew Hoe Tung

Author 4: Aroland Kiring

Keywords: Distress speech; ensemble bagged trees; audio surveillance; machine learning; distance; directions

Paper 32: A Feature-based Transfer Learning to Improve the Image Classification with Support Vector Machine

Abstract: In the big data era there are some issues regarding real-world classification problems. Some of the important challenges that still need to be overcome to produce an accurate classification model are the data imbalance, difficulties in labeling process, and differences on data distribution. Most classification problems are related to the differences in the data distribution and the lack of labels on some datasets while other datasets have abundant labels. To address the problem, this paper proposes a weighted-based feature-transfer learning (WbFTL) method to transfer knowledge between different but related domains, called cross-domain. The knowledge transfer is done through making a new feature representations in order to reduce the cross-domain’s distribution differences while maintaining the local structure of the domain. To make the new feature representation we implement a feature selection and inter-cluster class distance. We propose two stages of the feature selection process to capture the knowledge of the feature and its relation to the label. The first stage uses a threshold to select the feature. The second stage uses ANOVA (Analysis of Variance) to select features that are significant to the label. To enhance the accuracy, the selected features are weighted before being used for the training process using SVM. The proposed WbFTL are compared to 1-NN and PCA as baseline 1 and baseline 2. Both baseline models represent the traditional machine learning and dimensionality reduction method, without implementing transfer learning. It is also compared with TCA, the first feature-transfer learning work on this same task, as baseline 3. The experiment results of 12 cross-domain tasks on Office and Caltech dataset show that the proposed WbFTL can increase the average accuracy by 15.25%, 6.83%, and 3.59% compared to baseline 1, baseline 2, and baseline 3, respectively.

Author 1: Nina Sevani

Author 2: Kurniawati Azizah

Author 3: Wisnu Jatmiko

Keywords: Feature-transfer learning; image; feature selection; weight; distance

Paper 33: Distributed Training of Deep Autoencoder for Network Intrusion Detection

Abstract: The amount of data being exchanged over the internet is enormous. Attackers are finding novel ways to evade rules, investigate network defenses, and launch successful attacks. Intrusion detection is one of the effective means to counter attacks. As the network traffic continues to grow, it can be challenging for network administrators to detect intrusions. In huge networks connected with millions of computers Terabytes/Zettabytes of data is generated every second. Deep Learning is an effective means for analyzing network traffic and detecting intrusions. In this article, distributed autoencoder on the CSE-CIC-IDS2018 dataset is implemented by considering all the classes of the dataset. The proposed work is implemented on Azure Cloud using distributed training as it helps in speeding up the training process, thereby detecting intrusions faster. An overall accuracy of 98.96 % is achieved. By leveraging such parallel computing into the security process, organizations may accomplish operations more quickly and respond to risks and remediate them at a rate that would not be possible with manual human capabilities alone.

Author 1: Haripriya C

Author 2: Prabhudev Jagadeesh M. P

Keywords: Network intrusion detection systems; deep learning; autoencoders; cloud computing; distributed training; parallel computing

Paper 34: An Arabic Intelligent Diagnosis Assistant for Psychologists using Deep Learning

Abstract: Mental illnesses have increased in recent years, especially after Covid-19 pandemic. In Saudi Arabia, the number of psychiatric clinics is small compared to the population density. As a result, psychologists encounter a variety of difficulties at work. The main goal of the current research is to develop a system that assists psychologists in the diagnosis process, which will be based on the DSM-5 (Diagnosis and Statistical Manual of Mental Disorders). The work on this research started with collecting the requirements and identifying users’ needs. In this matter, several interviews have been conducted with Saudi Psychologist and then a questionnaire was developed and distributed to psychologists in Saudi Arabia. Following an analysis of the needs and requirements, the system was designed. A deep learning technique was applied during the diagnosing process to address the issues mentioned by psychologists. Additionally, the proposed system helps psychologists by quickly calculating the results of psychological tests. The system was built as a website. The Convolutional Neural Network (CNN) algorithm was used with 96% accuracy to automatically predict the appropriate diagnosis and suggest the most suitable psychological test for the patient to take. System testing and usability testing were also conducted by involving patients and Saudi psychologists to test the usability of the system and the accuracy of the CNN model. The results indicate that the diagnosis prediction was accurate, and that each activity was completed faster. This demonstrated the model's high degree of accuracy and the system's interfaces' clarity. Additionally, psychologists' comments were encouraging and positive.

Author 1: Asmaa Alayed

Author 2: Manar Alrabie

Author 3: Sarah Aldumaiji

Author 4: Ghaida Allhyani

Author 5: Sahar Siyam

Author 6: Reem Qaid

Keywords: Mental health; psychologist; mental illness diagnosis; psychological test; deep learning; CNN algorithm

Paper 35: The Evaluation of a Persuasive Learning Tool using Think-Aloud Protocol

Abstract: e-Learning has become a platform for students to gain and expand their knowledge through mobile applications or web-based systems. Even though e-learning systems usually aim to facilitate students' understanding of the subject, some fail to convey the underlying learning outcomes. These circumstances emerge as most e-learning methods or tools fail to attract students to engage in their studies continuously. Therefore, to overcome the problem, the Persuasive Learning Objects and Technologies (PLOT) model comprises persuasive design elements for online learning, is developed. A web-based statistical analysis assistant system called TemanKajianKu (Study Buddy) has been developed based on PLOT elements to assist students in identifying the correct approach to conduct and analyze their experiment. This paper aims to evaluate users’ experience and examine the effectiveness of the persuasive design elements of the system. Ten participants were involved in interviews using the Think-Aloud protocol method. The study results showed that most participants conveyed positive opinions by giving good feedback on the system design. Most also stated that the system could help them make decisions by utilizing persuasive elements such as reduction, social signal, tunnelling, tailoring, and self-monitoring. This concludes that the Persuasive Learning Tool is effective in helping develop an e-learning application or web-based system that helps students in decision-making concerning their studies.

Author 1: Muhammad Aqil Abd Rahman

Author 2: Mohamad Hidir Mhd Salim

Author 3: Nazlena Mohamad Ali

Keywords: Learning technology; persuasive technology; persuasive learning; persuasive design

Paper 36: Real-Time Intrusion Detection of Insider Threats in Industrial Control System Workstations Through File Integrity Monitoring

Abstract: Industrial control systems (ICS) play a crucial role in various industries and ensuring their security is paramount for maintaining process continuity and reliability. In ICS, the most damaging cyber-attacks often come from trusted insiders rather than external threats or malware. Insiders have the advantage of bypassing security measures and staying undetected. This research focuses on developing a real-time intrusion detection system for ICS workstations that effectively detects insider threats while prioritizing user privacy. The approach employs file integrity monitoring to identify suspicious activities, particularly file violations such as data tampering and destruction. The model presented in this research demonstrates low system resource consumption by utilizing an event-triggered approach instead of continuous polling of file data. The model leverages built-in operating system functions, eliminating the need for third-party software installation. To minimize disruptions to the ICS network, the model operates at the supervisory level within the ICS architecture. Through extensive testing, the model achieves a high level of accuracy, detecting insider intrusions with a high true positive rate. This reliable detection capability contributes to enhancing the security of ICS and mitigating the risks associated with insider threats. By implementing this real-time intrusion detection system, organizations can effectively protect their control systems while preserving user privacy.

Author 1: Bakil Al-Muntaser

Author 2: Mohamad Afendee Mohamed

Author 3: Ammar Yaseen Tuama

Keywords: Industrial control system; insider threats; intrusion detection; file integrity monitoring; SCADA security

Paper 37: Clustering Based on Gray Wolf Optimization Algorithm for Internet of Things over Wireless Nodes

Abstract: The Internet of Things (IoT) creates an environment where things are permitted to act, hear, listen, and talk. IoT devices encompass a wide range of objects, from basic sensors to intelligent devices, capable of exchanging information with or without human intervention. However, the integration of wireless nodes in IoT systems brings about both advantages and challenges. While wireless connectivity enhances system functionality, it also introduces constraints on resources, including power consumption, memory, and CPU processing capacity. Among these limitations, energy consumption emerges as a critical challenge. To address these challenges, metaheuristic algorithms have been widely employed to optimize routing patterns in IoT networks. This paper proposes a novel clustering strategy based on the Gray Wolf Optimization (GWO) algorithm. The GWO-based clustering approach aims to achieve energy efficiency and improve overall network performance. Experimental results demonstrate significant improvements in key performance metrics. Specifically, the proposed strategy achieves up to a 14% reduction in energy consumption, a 34% decrease in end-to-end delay, and a 10% increase in packet delivery rate compared to existing approaches. The findings of this research contribute to the advancement of energy-efficient and high-performance IoT networks. The utilization of the GWO algorithm for clustering enhances the network's ability to conserve energy, reduce latency, and improve the delivery of data packets. These outcomes highlight the effectiveness and potential of the proposed approach in addressing resource limitations and optimizing performance in IoT environments.

Author 1: Chunfen HU

Author 2: Haifei ZHOU

Author 3: Shiyun LV

Keywords: Internet of things; energy consumption; clustering; optimization; gray wolf optimization

Paper 38: Intelligent Moroccan License Plate Recognition System Based on YOLOv5 Build with Customized Dataset

Abstract: The rising number of automobiles has led to an increased demand for a reliable license plate identification system that can perform effectively in diverse conditions. This applies to local authorities, public organizations, and private companies in Morocco, as well as worldwide. To meet this need, a strong License Plate Recognition (LPR) system is required, taking into account local plate specifications and fonts used by plate manufacturers. This paper presents an intelligent LPR system based on the YOLOv5 framework, trained on a customized dataset encompassing multiple fonts and circumstances such as illumination, climate, and lighting. The system incorporates an intelligent region segmentation level that adapts to the plate's type, improving recognition accuracy and addressing separator issues. Remarkably, the model achieves an impressive precision rate of 99.16% on problematic plates with specific illumination, separators, and degradations. This research represents a significant advancement in the field of license plate recognition, providing a reliable solution for accurate identification and paving the way for broader applications in Morocco and beyond.

Author 1: El Mehdi Ben Laoula

Author 2: Marouane Midaoui

Author 3: Mohamed Youssfi

Author 4: Omar Bouattane

Keywords: License plate recognition; YOLOv5; intelligent region segmentation; customized dataset; Moroccan license plate issues; fonts-based data

Paper 39: Deep Learning for Personal Activity Recognition Under More Complex and Different Placement Positions of Smart Phone

Abstract: Personal Activity Recognition (PAR) is an indispensable research area as it is widely used in applications such as security, healthcare, gaming, surveillance and remote patient monitoring. With sensors introduced in smart phones, data collection for PAR made easy. However, PAR is non-trivial and difficult task due to bulk of data to be processed, complexity and sensor placement positions. Deep learning is found to be scalable and efficient in processing such data. However, the main problem with existing solutions is that, they could recognize up to 6 or 8 actions only. Besides, they suffer from accurate recognition of other actions and also deal with complexity and different placement positions of smart phone. To address this problem, in this paper, we proposed a framework named Robust Deep Personal Action Recognition Framework (RDPARF) which is based on enhanced Convolutional Neural Network (CNN) model which is trained to recognize 12 actions. RDPARF is realized with our proposed algorithm known as Enhanced CNN for Robust Personal Activity Recognition (ECNN-RPAR). This algorithm has provision for early stopping checkpoint to optimize resource consumption and faster convergence. Experiments are made with MHealth benchmark dataset collected from UCI repository. Our empirical results revealed that ECNN-RPAR could recognize 12 actions under more complex and different placement positions of smart phone besides outperforming the state of the art exhibiting highest accuracy with 96.25%.

Author 1: Bhagya Rekha Sangisetti

Author 2: Suresh Pabboju

Keywords: Human activity recognition; deep learning; CNN; MHealth dataset; artificial intelligence

Paper 40: A Review on Security Techniques in Image Steganography

Abstract: Given the increased popularity of the internet, the exchange of sensitive information leads to concerns about privacy and security. Techniques such as steganography and cryptography have been employed to protect sensitive information. Steganography is one of the promising tools for securely exchanging sensitive information through an unsecured medium. It is a powerful tool for protecting a user’s data, wherein the user can hide messages inside other media, such as images, videos, and audios (cover media). Image steganography is the science of concealing secret information inside an image using various techniques. The nature of the embedding process makes the hidden information undetectable to human eyes. The challenges faced by image steganography techniques include achieving high embedding capacity, good imperceptibility, and high security. These criteria are inter-related since enhancing one factor undermines one or more others. This paper provides an overview of existing research related to various techniques and security in image steganography. First, basic information in this domain is presented. Next, various kinds of security techniques used in steganography are explained, such as randomization, encryption, and region-based techniques. This paper covers research published from 2017 to 2022. This review is not exhaustive and aims to explore state-of-the-art techniques applied to enhance security, crucial issues in the domain, and future directions to assist new and current researchers.

Author 1: Sami Ghoul

Author 2: Rossilawati Sulaiman

Author 3: Zarina Shukur

Keywords: Image steganography; data hiding; steganographic security; randomization; encryption

Paper 41: Automated Type Identification and Size Measurement for Low-Voltage Metering Box Based on RGB-Depth Image

Abstract: The low-voltage metering box is a critical piece of equipment in the power supply system. The automated inspection of metering boxes is important in their production, transportation, installation, operation and maintenance. In this work, an automated type identification and size measurement method for low-voltage metering boxes based on RGB-D images is proposed. The critical components, including the door shell and window, connection terminal block, and metering compartment in the cabinet, are segmented first using the Mask-RCNN network. Then the proposed Sub-Region Closer-Neighbor algorithm is used to estimate the number of connection terminal blocks. Combined with the number of metering compartments, the type of metering box is classified. To refine the borders of the metering box components, an edge correction algorithm based on the Depth Difference (Dep-D) Constraint is presented. Finally, the automated size measurement is implemented based on the proposed Equal-Region Averaging algorithm. The experimental results show that the accuracies of the automated type identification and size measurement of the low-voltage metering box reach more than 92%.

Author 1: Pengyuan Liu

Author 2: Xurong Jin

Author 3: Shaokui Yan

Author 4: Tingting Hu

Author 5: Yuanfeng Zhou

Author 6: Ling He

Author 7: Xiaomei Yang

Keywords: Low-voltage metering box; RGB-D image processing; automated size detection; automated type detection; inspection automation

Paper 42: Data-driven Decision Making in Higher Education Institutions: State-of-play

Abstract: The paper highlights the importance of using data-driven decision-making tools in Higher Education Institutions (HEIs) to improve academic performance and support sustainable development. HEIs must utilize data analytics tools, including educational data mining, learning analytics, and business intelligence, to extract insights and knowledge from educational data. These tools can help HEIs’ leadership monitor and improve student enrolment campaigns, track student performance, evaluate academic staff, and make data-driven decisions. Although decision support systems have many advantages, they are still underutilized in HEIs, leaving field for further research and implementation. To address this, the authors summarize the benefits of applying data-driven decision approaches in HEIs and review various frameworks and methodologies, such as a course recommendation system and an academic prediction model, to aid educational decision-making. These tools articulate pedagogical theories, frameworks, and educational phenomena to establish mainstay significant components of learning to enable the scheming of superior learning systems. The tools can be utilized by the placement agencies or companies to find out their probable trainees/ recruitees. They can help students in course selection, and educational management in being more efficient and effective.

Author 1: Silvia Gaftandzhieva

Author 2: Sadiq Hussain

Author 3: Slavoljub Hilcenko

Author 4: Rositsa Doneva

Author 5: Kirina Boykova

Keywords: Business intelligence; data analytics tools; decision-making framework; decision-making support systems

Paper 43: Semi-Dense U-Net: A Novel U-Net Architecture for Face Detection

Abstract: Face detection and localization has been a major field of study in facial analysis and computer vision. Several convolutional neural network-based architectures have been proposed in the literature such as cascaded approach, single-stage and two-stage architectures. Using image segmentation based technique for object/face detection and recognition have been an alternative approach recently being employed. In this paper, we propose detection of faces by using U-net segmentation architectures. Motivated from DenseNet, a variant of U-net, called Semi-Dense U-Net, is designed in order to improve the binary masks generated by the segmentation model and further post-processed to detect faces. The proposed U-Net model have been trained and tested on FDDB, Wider face and Open Image dataset and compared with state-of-the-art algorithms. We could successfully achieve dice coefficient of 95.68% and average precision of 91.60% on a set of test data from OpenImage dataset.

Author 1: Ganesh Pai

Author 2: Sharmila Kumari M

Keywords: Semi-Dense U-Net; face detection; segmentation; U-Net

Paper 44: End to End Text to Speech Synthesis for Malay Language using Tacotron and Tacotron 2

Abstract: Text-to-speech (TTS) technology is becoming increasingly popular in various fields such as education and business. However, the advancement of TTS technology for Malay language is slower compared to other language especially English language. The rise of artificial intelligence (AI) technology has sparked TTS technology into a new dimension. An end-to-end (E2E) TTS system that generates speech directly from text input is one of the latest AI technologies for TTS and implementing this E2E method into Malay language will help to expand the TTS technology for Malay language. This study involves the development and comparison of two end-to-end TTS models for the Malay language, namely Tacotron and Tacotron 2. Both models were trained using a Malay corpus consisting of text and speech and evaluated the synthesized speech using Mean Opinion Scores (MOS) for naturalness and intelligibility. The results show that Tacotron outperformed Tacotron 2 in terms of naturalness and intelligibility, with both models falling short of human speech quality. Improving TTS technology for Malay can encourage its use in a wider range of contexts.

Author 1: Azrul Fahmi Abdul Aziz

Author 2: Sabrina Tiun

Author 3: Noraini Ruslan

Keywords: Text to speech; end-to-end TTS; Tacotron; Tacotron 2; Malay language; artificial intelligence; mean opinion score (MOS); naturalness; intelligibility

Paper 45: A New Model for Blood Cancer Classification Based on Deep Learning Techniques

Abstract: Artificial intelligence and deep learning algorithms have become essential fields in medical science. These algorithms help doctors detect diseases early, reduce the incidence of errors, and decrease the time required for disease diagnosis, thereby saving human lives. Deep learning models are widely used in Computer-Aided Diagnosis Systems (CAD) for the classification of various diseases, including blood cancer. Early diagnosis of blood cancer is crucial for effective treatment and saving patients' lives. Therefore, this study developed two distinct models to classify eight types of blood cancer. These types include follicular lymphoma (FL), mantle cell lymphoma (MCL), chronic lymphocytic leukemia (CLL), acute myeloid leukemia (AML), and the subtypes of acute lymphoblastic leukemia (ALL) known as early pre-B, pre-B, pro-B ALL, and benign. AML and ALL are specific classifications for human leukemia cancer, while FL, MCL, and CLL are specific classifications for lymphoma. Both models consist of different phases, including data collection, preprocessing, feature extraction techniques, and the classification process. The techniques applied in these phases are the same in both proposed models, except for the classification phase. The first model utilizes the VGG16 architecture, while the second model utilizes DenseNet-121. The results indicated that DenseNet-121 achieved a lower accuracy compared to VGG16. VGG16 exhibited excellent results, achieving an accuracy of 98.2% when classifying the eight classes. This outcome suggests that VGG16 is the most effective classifier for the utilized dataset.

Author 1: Hagar Mohamed

Author 2: Fahad Kamal Elsheref

Author 3: Shrouk Reda Kamal

Keywords: Deep learning; convolutional neural networks (CNNs); leukemia; lymphoma; computer-aided diagnosis systems (CAD)

Paper 46: Deep Feature Fusion Network for Lane Line Segmentation in Urban Traffic Scenes

Abstract: As autonomous driving technology continues to advance at a rapid pace, the demand for precise and dependable lane detection systems has become increasingly critical. However, traditional methods often struggle with complex urban scenarios, such as crowded environments, diverse lighting conditions, unmarked lanes, curved lanes, and night-time driving. This paper presents a novel approach to lane line segmentation in urban traffic scenes with a Deep Feature Fusion Network (DFFN). The DFFN leverages the strengths of deep learning for feature extraction and fusion, aiming to enhance the accuracy and reliability of lane detection under diverse real-world conditions. To integrate multi-layer features, the DFFN employs both spatial and channel attention mechanisms in an appropriate manner. This strategy facilitates learning and predicting the relevance of each input feature during the fusion process. In addition, deformable convolution is employed in all up-sampling operations, enabling dynamic adjustment of the receptive field according to object scales and poses. The performance of DFFN is rigorously evaluated and compared with existing models, namely SCNN, ENet, and ENet-SAD, across different scenarios in the CULane dataset. Experimental results demonstrate the superior performance of DFFN across all conditions, highlighting its potential applicability in advanced driver assistance systems and autonomous driving applications.

Author 1: Hoanh Nguyen

Keywords: Lane line segmentation; deep learning; convolutional neural network; spatial and channel attention

Paper 47: Enhancing Skin Diseases Classification Through Dual Ensemble Learning and Pre-trained CNNs

Abstract: Skin diseases represent a variety of disorders that can affect the skin. In fact, early diagnosis plays a central role in the treatment of this type of disease. This scholarly article introduces a novel approach to classifying skin diseases by leveraging two ensemble learning techniques, encompassing multi-modal and multi-task methodologies. The proposed classifier integrates diverse information sources, including skin lesion images and patient-specific data, aiming to enhance the accuracy of disease classification. By simultaneously utilizing image input and structured data input, the multi-task functionality of the classifier enables efficient disease classification. The integration of multi-modal and multi-task techniques allows for a comprehensive analysis of skin diseases, leading to improved classification performance and a more holistic understanding of the underlying factors influencing disease diagnosis. The efficacy of the classifier was assessed using the ISIC 2018 dataset, which comprises both image and clinical information for each patient with skin diseases. The dataset used in this study comprises images of seven different types of skin diseases and their associated medical information. The findings of our proposed approach show that it outperforms traditional single-modal and single-task classifiers. The results of this study demonstrate that the proposed model attained an accuracy of 97.66% for the initial classification task (image classification). Additionally, the second classification task (clinical data classification) achieved an accuracy of 94.40%.

Author 1: Oussama El Gannour

Author 2: Soufiane Hamida

Author 3: Yasser Lamalem

Author 4: Bouchaib Cherradi

Author 5: Shawki Saleh

Author 6: Abdelhadi Raihani

Keywords: Multi-modal approach; multi-task approach; transfer learning; deep learning; skin diseases classification

Paper 48: Auto-Regressive Integrated Moving Average Threshold Influence Techniques for Stock Data Analysis

Abstract: This study focuses on predicting and estimating possible stock assets in a favorable real-time scenario for financial markets without the involvement of outside brokers about broadcast-based trading using various performance factors and data metrics. Sample data from the Y-finance sector was assembled using API-based data series and was quite accurate and precise. Prestigious machine learning algorithmic performances for both classification and regression complexities intensify this assumption. The fallibility of stock movement leads to the production of noise and vulnerability that relate to decision-making. In earlier research investigations, fewer performance metrics were used. In this study, Dickey-Fuller testing scenarios were combined with time series volatility forecasting and the Long Short-Term Memory algorithm, which was used in a futuristic recurrent neural network setting to predict future closing prices for large businesses on the stock market. In order to analyze the root mean squared error, mean squared error, mean absolute percentage error, mean deviation, and mean absolute error, this study combined LSTM methods with ARIMA. With fewer hardware resources, the experimental scenarios were framed, and test case simulations carried out.

Author 1: Bhupinder Singh

Author 2: Santosh Kumar Henge

Author 3: Sanjeev Kumar Mandal

Author 4: Manoj Kumar Yadav

Author 5: Poonam Tomar Yadav

Author 6: Aditya Upadhyay

Author 7: Srinivasan Iyer

Author 8: Rajkumar A Gupta

Keywords: Dickey-Fuller test case (DF-TC); recurrent neural network (RNN); root mean square error (RMSE); long short-term memory (LSTM); machine learning (ML); auto-regressive integrated moving average (ARIMA)

Paper 49: Evaluations on Competitiveness of Service Sector in Yangtze River Economic Belt of China Based on Dual-core Diamond Model

Abstract: By expanding and innovating Michael Porter Diamond Model, a Dual-core Diamond Model is developed in this paper with innovation and openness as the core factors in consideration of the actual needs of the development of service sector in the Yangtze River Economic Belt of China. This paper establishes an evaluation indicator system of service sector competitiveness profitability to measure and evaluate the competitiveness of service sector in 11 provinces and towns in the Yangtze River Economic Belt through PCA (principal component analysis) based on the relevant information of the 11 provinces and towns mentioned above in 2015 and 2016. The research results indicate that the design of Dual-core Diamond Model is in line with the current situation and future development needs of the service sector in the Yangtze River Economic Belt, and the dual-core factors, namely, innovation and openness, have become the most important factors influencing the competitiveness of the service sector in the Yangtze River Economic Belt. Based on the model analysis results, it should propose strategies to enhance the competitiveness of the service industry in the Yangtze River Economic Belt. It needs to enhance innovation ability as well as to further expand trade in services. Firstly, encourage the growth of related industries and create a coordinated development cluster for the service sector. Second, intensify efforts in talent cultivation and build a talent system in alignment with the development of service sector. Third, improve the relevant legal system and innovate the service supervision and governance system in the service sector. Last, focus on a coordinated and integrated inter-region development.

Author 1: Ming Zhao

Author 2: Qingjun Zeng

Author 3: Dan Wang

Author 4: Jiafu Su

Keywords: Dual-core diamond model; service sector; Yangtze river economic belt; principal component

Paper 50: Enhancing COVID-19 Diagnosis Through a Hybrid CNN and Gray Wolf Optimizer Framework

Abstract: Covid-19 is an infectious respiratory disorder brought about using a brand-new coronavirus first found in 2019. The severity of symptoms can vary from mild to life-threatening. No vaccine or specific treatment has been developed to address Covid-19. Hence the most effective preventive measure is to practice social distancing and adhere to good hygiene practices. Medical imaging and convolutional neural networks are used in Covid-19 research to quickly identify infected individuals and detect changes in the lung tissue of those infected. Convolutional neural networks can be used to analyze chest CT scans, detecting potential signs of infection like ground-glass opacities, which indicate the presence of Covid-19. This article introduces a powerful framework for classifying COVID-19 images utilizing a hybrid of CNN and an improved version of Gray Wolf Optimizer. To demonstrate the efficiency of the projected framework, it is verified on a standard dataset and compared with other methods, with results indicating its superiority over the others.

Author 1: Yechun JIN

Author 2: Guanxiong ZHANG

Author 3: Jie LI

Keywords: Covid-19; respiratory disorder; medical imaging; convolutional neural networks; improved gray wolf algorithm

Paper 51: Automated Epileptic Seizure Detection using Improved Crystal Structure Algorithm with Stacked Autoencoder

Abstract: Epilepsy can be referred to as a neurological disorder, categorized by intractable seizures with serious consequences. To forecast such seizures, Electroencephalogram (EEG) datasets should be gathered continuously. EEG signals were recorded by using numerous electrodes fixed on the scalp that cannot be worn by patients continuously. Neurostimulators can intervene in advance and ignore the seizure rate. Its productivity is increased by using heuristics such as advanced seizure prediction. In recent times, several authors have deployed various deep learning approaches for predicting epileptic seizures, utilizing EEG signals. In this work, an Automated Epileptic Seizure Detection using Improved Crystal Structure Algorithm with Stacked Auto encoder (AESD-ICSASAE) technique has been developed. The presented AESD-ICSASAE technique executes a three-stage process. At the initial level, the AESD-ICSASAE technique applies min-max normalization approach to normalize the input data. Next, the AESD-ICSASAE technique uses ICSA based feature selection method for optimal choice of features. Finally, the SAE based classification process takes place and the hyperparameter selection process is performed by Arithmetic Optimization Algorithm (AOA). To depict the enhanced classification outcomes of the AESD-ICSASAE technique, series of experiments was made. Furthermore, the proposed method's results have been tested utilizing the CHB-MIT database, with results indicating an accuracy of 98.9%. These results validate the highest level of accuracy in seizure classification across all of the analyzed EEG data. A full set of experiments validated the AESD-ICSASAE method's enhancements.

Author 1: Srikanth Cherukuvada

Author 2: R. Kayalvizhi

Keywords: Deep learning; EEG signals; epileptic seizure detection; hyperparameter tuning; stacked autoencoders

Paper 52: Evaluation of the Effects of 2D Animation on Business Law: Elements of a Valid Contract

Abstract: This article presents an evaluation of Business Law 2D Animation: Elements of a Valid Contract. The developed application was produced to assist business law students to understand the contents of the topic Elements of a Valid Contract. An experiment was carried out to assess the usability of the application as a learning and review material for business law students. This study comprised five major evaluation components, including learnability, usability, accessibility, functionality, and effectiveness, to investigate user involvement and satisfaction with the proposed educational learning system. To acquire user testing results, online questionnaires were issued. There was a total of 63 respondents, including multimedia experts, students, and subject matter experts. The findings of the current study revealed that the majority of respondents were pleased with the outcomes of the animation. The results may assist in improving the teaching of the topic Elements of a Valid Contract for business law students as it provides visually appealing method of learning.

Author 1: Sarni Suhaila Rahim

Author 2: Nur Zulaiha Fadlan Faizal

Author 3: Shahril Parumo

Author 4: Hazira Saleh

Keywords: 2D Animation; business law; elements of a valid contract; evaluation

Paper 53: A Novel Approach to Multi-Layer-Perceptron Training using Quadratic Interpolation Flower Pollination Neural Network on Non-Binary Datasets

Abstract: Machine Learning (ML) algorithms are widely used in solving classification problems. The biggest challenge of classification lies in the robustness of the ML algorithm in various dataset characteristics. Quadratic Interpolation Flower Pollination Neural Network (QIFPNN) is categorised into ML algorithm. The new QIFPNN's extraordinary capabilities are measured on binary-type datasets. This research ensures that the remarkable ability of QIFPNN also applies to non-binary datasets with balanced and unbalanced data class characteristics. Flower Pollination Neural Network (FPNN), Particle Swarm Optimisation Neural Network (PSONN), and Bat Neural Network (BANN) were used as comparisons. The QIFPNN, FPNN, PSONN, and BANN were used to train Multi-Layer-Perceptron (MLP). The test results on five datasets show that QIFPNN obtains an average classification accuracy higher than its comparison in three datasets with balanced and unbalanced data class characteristics. The three datasets are Iris, Wine, and Glass. The highest classification accuracy obtained by QIFPNN in the three datasets is 97.1462%, 98.6551%, and 73.1979%, respectively. Based on the F1-score test from QIFPNN, it is higher than all the comparisons in four datasets: Iris, Wine, Vertebral column, and Glass. Sequentially, 96.4599%, 98.7155%, 90.7517%, and 60.2843%. It proves that QIFPNN can also classify datasets with non-binary data types with balanced and unbalanced data class characteristics because they are more consistently tested on various datasets and are not susceptible to the influence of variations in dataset characteristics so that they can be applied to various types of data or cases.

Author 1: Yulianto Triwahyuadi Polly

Author 2: Sri Hartati

Author 3: Suprapto

Author 4: Bambang Sumiarto

Keywords: Quadratic interpolation; flower pollination algorithm; neural network; non-binary dataset; multi-layer-perceptron

Paper 54: Hamming Distance Approach to Reduce Role Mining Scalability

Abstract: Role-based Access Control has become the standard of practice for many organizations for restricting control on limited resources in complicated infrastructures or systems. The main objective of the role mining development is to define appropriate roles that can be applied to the specified security access policies. However, the mining scales in this kind of setting are extensive and can cause a huge load on the management of the systems. To resolve the above mentioned problems, this paper proposes a model that implements Hamming Distance approach by rearranging the existing matrix as the input data to overcome the scalability problem. The findings of the model show that the generated file size of all datasets substantially have been reduced compared to the original datasets It has also shown that Hamming Distance technique can successfully reduce the mining scale of datasets ranging between 30% and 47% and produce better candidate roles.

Author 1: Nazirah Abd Hamid

Author 2: Siti Rahayu Selamat

Author 3: Rabiah Ahmad

Author 4: Mumtazimah Mohamad

Keywords: Role-based Access Control; role mining; hamming distance; data mining

Paper 55: Towards Path Planning Algorithm Combining with A-Star Algorithm and Dynamic Window Approach Algorithm

Abstract: In the Automated Guided Vehicle (AGV) warehouse automatic guided vehicle system, the path planning algorithm for intelligent logistics vehicles is a key factor to ensure the stable and efficient operation of the system. However, the existing planning algorithms have problems such as single designing a route and the inability to intelligently evade moving barriers. The academic community has proposed various solutions to these problems, although they have improved the efficiency and quality of path planning to some extent, they have not completely solved problems such as poor safety in planning, the high number of path inflection points, poor path smoothness, easily getting stuck in deadlocks, and have not fully considered the running cost and practical implementation difficulty of algorithms. To address these issues, the article deeply researched traditional A* scheme and Dynamic Window Approach (DWA) technology and proposed designing a route method according to the fusion of the A* algorithm and DWA technology. The algorithm improved the A algorithm by introducing a sub-node optimization algorithm to solve problems for instance poor global path planning safety and easy deadlock. Moreover, the algorithm reduced the amount of global route reversal locations and increased path consistency by improving the evaluation function and removing redundant points of the A algorithm. Finally, by integrating the DWA algorithm, the intelligent logistics vehicle achieved dynamic obstacle avoidance capabilities for moving objects in the real world. Our simulations-based results on MATLAB framework show that the algorithm significantly improves path smoothness, path length, path planning time, and environmental adaptability compared to traditional algorithms, and basically meets the path planning requirements of the AGV system for intelligent logistics vehicles.

Author 1: Kaiyu Li

Author 2: Xiugang Gong

Author 3: Muhammad Tahir

Author 4: Tao Wang

Author 5: Rajesh Kumar

Keywords: AGV; path planning; A* algorithm; dynamic window approach

Paper 56: Advances in Machine Learning and Explainable Artificial Intelligence for Depression Prediction

Abstract: There is a growing interest in applying AI technology in the field of mental health, particularly as an alternative to complement the limitations of human analysis, judgment, and accessibility in mental health assessments and treatments. The current mental health treatment service faces a gap in which individuals who need help are not receiving it due to negative perceptions of mental health treatment, lack of professional manpower, and physical accessibility limitations. To overcome these difficulties, there is a growing need for a new approach, and AI technology is being explored as a potential solution. Explainable artificial intelligence (X-AI) with both accuracy and interpretability technology can help improve the accuracy of expert decision-making, increase the accessibility of mental health services, and solve the psychological problems of high-risk groups of depression. In this review, we examine the current use of X-AI technology in mental health assessments for depression. As a result of reviewing 6 studies that used X-AI to discriminate high-risk groups of depression, various algorithms such as SHAP (SHapley Additive exPlanations) and Local Interpretable Model-Agnostic Explanation (LIME) were used for predicting depression. In the field of psychiatry, such as predicting depression, it is crucial to ensure AI prediction justifications are clear and transparent. Therefore, ensuring interpretability of AI models will be important in future research.

Author 1: Haewon Byeon

Keywords: Depression; LIME; Explainable artificial intelligence; Machine learning; SHAP

Paper 57: State of-the-Art Analysis of Multiple Object Detection Techniques using Deep Learning

Abstract: Object detection has experienced a surge in interest due to its relevance in video analysis and image interpretation. Traditional object detection approaches relied on handcrafted features and shallow trainable algorithms, which limited their performance. However, the advancement of Deep learning (DL) has provided more powerful tools that can extract semantic, high- level, and deep features, addressing the shortcomings of previous systems. Deep Learning-based object detection models differ regarding network architecture, training techniques, and optimization functions. In this study, common generic designs for object detection and various modifications and tips to enhance detection performance have been investigated. Furthermore, future directions in object detection research, including advancements in Neural Network-based learning systems and the challenges have been discussed. In addition, comparative analysis based on performance parameters of various versions of YOLO approach for multiple object detection has been presented.

Author 1: Kanhaiya Sharma

Author 2: Sandeep Singh Rawat

Author 3: Deepak Parashar

Author 4: Shivam Sharma

Keywords: Deep learning; neural networks; object detection; YOLO

Paper 58: Enhancing IoT Security with Deep Stack Encoder using Various Optimizers for Botnet Attack Prediction

Abstract: The Internet of Things (IoT) connects different sensors, devices, applications, databases, services, and people, bringing improvements to various aspects of our lives, such as cities, agriculture, finance, and healthcare. However, guaranteeing the safety and confidentiality of IoT data which has become rich in its quality requires careful preparation and awareness. Machine learning techniques are used to predict different types of cyber-attacks, including denial of service (DoS), botnet attacks, malicious operations, unauthorized control, data probing, surveillance, scanning, and incorrect setups. In this study, for improving security of IoT data, a method called Deep Stack Encoder Neural Network to predict botnet attacks by using N-BaIoT bench mark dataset is employed. In this study a new framework is introduced which will improve the performance of prediction rate to 94.5%. To evaluate the performance of this method assessment criteria are adopted like accuracy, precision, recall, and F1 score, comparing it with other models. From the optimizers of Adam, Adagrad and Adadelta, Adam optimizer gave the highest accuracy with relu activation function.

Author 1: Archana Kalidindi

Author 2: Mahesh Babu Arrama

Keywords: Internet of things; botnet attacks; neural network methods; N-BaIoT; deep stack encoder; Adam optimizer; Adagrad optimizer; Adadelta optimizer; activation function

Paper 59: Behavior Intention of Chronic Illness Patients in Malaysia to Use IoT-based Healthcare Services

Abstract: The Internet of Things (IoT) has emerged as a trend in the healthcare industry to develop innovative solutions that enhance patient outcomes and operational efficiency. Healthcare has become more accessible, affordable, and efficient to sensors, wearables, and health monitors. The healthcare industry's adoption of the Internet of Things is lagging behind other sectors despite its many benefits. This study aims to investigate the extent to which chronic patients in Malaysia are using healthcare services made possible by the Internet of Things. To that end, this study proposes a unified framework to examine how these highlighted factors affect Behavioral Intention (BI) with regard to adopting IoT healthcare services. The innovation here is in bringing together three distinct theories: i) the Technology-Organization-Environment Framework (TOE), which is a framework for understanding how companies adopt new technologies; ii) the Unified Theory of Acceptance and Use of Technology (UTAUT); and iii) the Social Exchange Theory (SE). Patients in Malaysia who are coping with long-term health issues were surveyed online. This study also employs SPSS and Smart Partial Least Square (Smart PLS) for data analysis. Eleven hypothesized predictive components have been investigated. The results showed that chronic illness patients' BI towards adopting IoT solutions was considerably impacted by both individual and technological factors and related aspects. The impact of BI on Use Behaviour (UB) also showed similar outcomes. Moreover, trust somewhat mediates the impact of both individual and technological factors on BI. The findings of this investigation will be beneficial to policymakers and suppliers of healthcare in that country. Additionally, the patients and their family members would gain benefits from the study due to the fact that the delivery of comprehensive treatment, especially in the field of chronic disease management, will be improved through IoT-healthcare services. The Internet of Things will also let medical staff function remotely and professionally.

Author 1: Huda Hussein Mohamad Jawad

Author 2: Zainuddin Bin Hassan

Author 3: Bilal Bahaa Zaidan

Keywords: Internet of things; IoT; chronic disease; adoption theories; adoption

Paper 60: Dynamic Difficulty Adjustment of Serious-Game Based on Synthetic Fog using Activity Theory Model

Abstract: This study used the activity theory model to determine the dynamic difficulty adjustment of serious-game based on synthetic fog. The difference in difficulty levels was generated in a 3-dimensional game environment with changes determined by applying varying fog thickness. The activity theory model in serious-games aims to facilitate development analysis in terms of learning content, the equipment used, and the resulting in-game action. The difficulty levels vary according to the player's ability because the game is expected to reduce boredom and frustration. Furthermore, this study simulated scenarios of various conditions, scores, time remaining, and the lives of synthetic players. The experimental results showed that the system can change the game environment with different fog thicknesses according to synthetic player parameters.

Author 1: Fresy Nugroho

Author 2: Puspa Miladin Nuraida Safitri Abdul Basid

Author 3: Firma Sahrul Bahtiar

Author 4: I. G. P. Asto Buditjahjanto

Keywords: Dynamic difficulty adjustment; serious-game; activity theory model; synthetic fog; synthetic player

Paper 61: Detection of Breast Cancer using Convolutional Neural Networks with Learning Transfer Mechanisms

Abstract: Breast cancer is the leading cause of mortality in women worldwide. One of the biggest challenges for physicians and technological support systems is early detection, because it is easier to treat and establish curative treatments. Currently, assistive technology systems use images to detect patterns of behavior with respect to patients who have been found to have some type of cancer. This work aims to identify and classify breast cancer using deep learning models and convolutional neural networks (CNN) with transfer learning. For the breast cancer detection process, 7803 real images with benign and malignant labels were used, which were provided by BreaKHis on the Kaggle platform. The convolutional basis (parameters) of pre-trained models VGG16, VGG19, Resnet-50 and Inception-V3 were used. The TensorFlow framework, keras and Python libraries were also used to retrain the parameters of the models proposed for this study. Metrics such as accuracy, error ratio, precision, recall and f1-score were used to evaluate the models. The results show that the models based on VGG16, VGG19 ResNet-50 and Inception-V3 obtain an accuracy of 88%, 86%, 97% and 96% respectively, recall of 84%, 82%, 96% and 96% respectively, in addition to f1-score of 86%, 83%, 96% and 95% respectively. It is concluded that the model that shows the best results is Resnet-50, obtaining high results in all the metrics considered, although it should be noted that the Inception-V3 model achieves very similar results in relation to Resnet-50, in all the metrics. In addition, these two models exceed the 95% threshold of correct results.

Author 1: Victor Guevara-Ponce

Author 2: Ofelia Roque-Paredes

Author 3: Carlos Zerga-Morales

Author 4: Andrea Flores-Huerta

Author 5: Mario Aymerich-Lau

Author 6: Orlando Iparraguirre-Villanueva

Keywords: Convolutional neural networks; transfer learning; deep learning; classification; breast cancer

Paper 62: Zero-Watermarking for Medical Images Based on Regions of Interest Detection using K-Means Clustering and Discrete Fourier Transform

Abstract: Watermarking schemes ensure digital image security and copyright protection to prevent unauthorized distribution. Zero-watermarking methods do not modify the image. This characteristic is a requirement in some tasks that need image integrity, such as medical images. Zero-watermarking methods obtain specific features for the master share construction to protect the digital image. This paper proposed a zero-watermarking scheme based on K-means clustering for ROI detection to obtain specific features. The K-means algorithm classifies the data according to the proximity of the generated clusters. K-means clustering is applied for image segmentation to identify ROI and detect areas that contain important information from the image. Therefore, the Discrete Fourier Transform (DFT) is applied to the ROI features, using the high frequencies to increase its robustness against geometric attacks. In addition, an edge detection based on the Sobel operator is applied for the QR code creation. This type of watermark avoids errors in watermark detection and increases the robustness of the watermark system. The master share creation is based on an XOR logic operation between extracted features from the selected ROI and the watermark. This method focuses on the protection of the image despite it being tampered with. Many proposed schemes focus on protection against advanced image processing attacks. The experiments demonstrate that the presented algorithm is robust against geometric and advanced signal-processing attacks. The DFT coefficients from the extracted ROI features increase the efficiency and robustness.

Author 1: Rodrigo Eduardo Arevalo-Ancona

Author 2: Manuel Cedillo-Hernandez

Keywords: Zero-watermark; ROI detection; machine learning; k-means; image security; copyright protection

Paper 63: Kalman Filter-based Signal Processing for Robot Target Tracking

Abstract: In the field of computer vision, the signal tracking of moving objects is a highly representative problem. Therefore, how to accurately and quickly track the target unit has become the focus of the research. Based on this, a Cam Shift algorithm improved by Kalman filtering algorithm is introduced to realize fast tracking of moving targets. This method uses the prediction function of the Kalman filter to predict the moving target of the next frame, transforms the global search problem into a local search problem, and improves the real-time performance. The experimental results show that, in the case of complete occlusion, the trajectory of the unimproved algorithm will deviate compared with the actual trajectory of the improved trajectory tracking curve, but the improved algorithm has no trajectory deviation. The error of the improved algorithm is about 4%, while the maximum error of the unimproved algorithm is about 90%. The improved algorithm reached the expected target accuracy after 110 and 78 trainings in X and Y coordinates, respectively, while the CamShift algorithm without Kalman filtering still failed to reach the expected error after 200 trainings in X and Y coordinates. This indicates that the performance of the improved CamShift algorithm based on Kalman filter has been greatly improved. In conclusion, the improved algorithm proposed in this study is highly practical.

Author 1: Baofu Gong

Keywords: Motion target tracking; Kalman filter; CamShift algorithm; occlusion processing

Paper 64: Vehicle Path Planning Based on Gradient Statistical Mutation Quantum Genetic Algorithm

Abstract: In the field of vehicle path planning, traditional intelligent optimization algorithms have the disadvantages of slow convergence, poor stability and a tendency to fall into local extremes. Therefore, a gradient statistical mutation quantum genetic algorithm (GSM-QGA) is proposed. Based on the dynamic rotation angle adjustment by the chromosome fitness value, the quantum rotation gate adjustment strategy is improved by introducing the idea of gradient descent. According to the statistical properties of chromosomal change trends, the gradient-based mutation operator is designed to realize the mutation operation. The shortest path is used as the metric to build the vehicle path planning model, and the effectiveness of the modified algorithm in vehicle path planning is demonstrated by simulation experiments. Compared with other optimization algorithms, the path length planned by the improved algorithm is shorter and the search stability is better. The algorithm can be effectively controlled to fall into local optimums.

Author 1: Hui Li

Author 2: Huiping Qin

Author 3: Zi’ao Han

Author 4: Kai Lu

Keywords: Quantum genetic algorithm; path planning; gradient descent; adaptive mutation operator; quantum rotation gate

Paper 65: Apache Spark in Healthcare: Advancing Data-Driven Innovations and Better Patient Care

Abstract: The enormous amounts of data produced in the healthcare sector are managed and analyzed with the help of Apache Spark, an open-source distributed computing system. This case study examines how Spark is utilized in the healthcare industry to produce data-driven innovations and enhance patient care. The report gives a general introduction of Spark's architecture, advantages, and healthcare use cases, such as managing electronic health records, predictive analytics for disease outbreaks, individualized medicine, medical image analysis, and remote patient monitoring. Additionally, it contains several case studies that highlight Spark's effects on lowering hospital readmission rates, detecting sepsis earlier, enhancing cancer research and therapy, and speeding up drug discovery. The report also identifies obstacles with data security and privacy, scalability and infrastructure, data integration and quality, labor and skills shortages, and other aspects of employing Spark in healthcare. Spark has overcome these obstacles by enabling efficient data-driven decision-making processes and enhancing patient outcomes, revolutionizing healthcare solutions. Additionally, the study looks at potential future advancements in healthcare, including the use of Spark with AI and ML, real-time analytics, the Internet of Medical Things (IoMT), enhanced interoperability and data sharing, and ethical standards. In conclusion, healthcare businesses can fully utilize Spark to transform their data into actionable insights that will enhance patient care and boost the efficiency of healthcare systems.

Author 1: Lalit Shrotriya

Author 2: Kanhaiya Sharma

Author 3: Deepak Parashar

Author 4: Kushagra Mishra

Author 5: Sandeep Singh Rawat

Author 6: Harsh Pagare

Keywords: Apache spark; healthcare; patient; styling; predictive analysis

Paper 66: Weight Optimization Based on Firefly Algorithm for Analogy-based Effort Estimation

Abstract: Proper cost estimation is one of the vital tasks that must be achieved for software project development. Owing to the complexity and uncertainties of the software development process, this task is ambiguous and difficult. Recently, analogy-based estimation (ABE) has become one of the popular approaches in this field due to its effectiveness and practicability in comparing completed projects and new projects in estimating the development effort. However, in spite of its many achievements, this method is not capable to guarantee accurate estimation confronting the complex relation between independent features and software effort. In such a case, the performance of the ABE can be improved by efficient feature weighting. This study introduces an enhanced software estimation method by integrating the firefly algorithm (FA) with the ABE method for improving software development effort estimation (SDEE). The proposed model can provide accurate identification of similar projects by optimising the performances of the similarity function in the estimation process in which the most relevant weights are assigned to project features for obtaining the more accurate estimates. A series of experiments were carried out using six real-world datasets. The results based on the statistical analysis showed that the integration of the FA and ABE significantly outperformed the existing analogy-based approaches especially for the ISBSG dataset.

Author 1: Ayman Jalal AlMutlaq

Author 2: Dayang N. A. Jawawi

Author 3: Adila Firdaus Binti Arbain

Keywords: Analogy-based estimation; firefly algorithm; software cost estimation; weight optimization

Paper 67: Hierarchical and Efficient Identity-bsed Encryption Against Side Channel Attacks

Abstract: Hierarchical and identity-based encryption (HIBE) is very valuable and widely used in many occasions. In the Internet of Things based on cloud services, efficient HIBE is likely to be applied to cloud service scenarios for the limited computing ability of some terminal devices. What’s more, because of the insecurity of cryptographic systems caused by side channel attacks, the design of leakage resilient cryptographic scheme has attracted more and more cryptography researchers' attention. In this study, an efficient leakage resilient HIBE is constructed. (1) In essence, this given scheme contains a hierarchical ID-based key encapsulation system. By using the extractor to act on the encapsulated symmetric key, this proposed scheme may resist the disclosure for the symmetric key due to side channel attacks. The relative leakage ratio of the encapsulated key is close to 1. (2) We also construct a hierarchical identity-based hash proof system that provides the security of our scheme. The proposed scheme can not only resist side channel attacks, but also has short public key parameters and computational efficiency, which is very suitable for applications in the Internet of Things environment. (3) There is no limit to the hierarchy depth of the system, and only the maximum hierarchy length is required to be given when the system is initialized.

Author 1: Qihong Yu

Author 2: Jian Shen

Author 3: Jiguo Li

Author 4: Sai Ji

Keywords: Identity-based encryption; side channel attack; hash proof system; composite order group

Paper 68: Image Specular Highlight Removal using Generative Adversarial Network and Enhanced Grey Wolf Optimization Technique

Abstract: Image highlight plays a major role in different interactive media and computer vision technology such as image fragmentation, recognition and matching. The original data will be unclear if the image contains highlights. Moreover, it may reduce the robustness in non-transparent as well as glassy objects and also it reduces accuracy. Hence, the removal of highlights is an extremely crucial thing in the dome of digital image enhancement. This is to develop the enhancement of the texture in imageries, and video analytics. Several state-of-art methods are used for removing highlights; but they face some difficulties like insufficient efficacy, accuracy and producing less datasets. To overcome this issue, this paper proposes an optimized GAN technology. The Enhanced Grey Wolf Optimization (EGWO) technique is employed for feature selection process. Generative Adversarial Network is a machine learning (ML) algorithm. Here, two neural networks that will compete among themselves to produce better calculations. The algorithm generates realistic data, especially images, with great practical results. The investigational outcome reveals that the future algorithm has the ability to verify and eliminate the illumination spotlight in the image so that real details can be obtained from the image. The effectiveness of the proposed work can be proved by comparing the proposed optimized GAN with other existing models in highlight removal task. The comparison outcome gives better accuracy with 99.91% compared to previous existing methods.

Author 1: Maddikera Krishna Reddy

Author 2: J. C. Sekhar

Author 3: Vuda Sreenivasa Rao

Author 4: Mohammed Saleh Al Ansari

Author 5: Yousef A.Baker El-Ebiary

Author 6: Jarubula Ramu

Author 7: R. Manikandan

Keywords: Highlight detection; optimization; specular highlight detection; GAN

Paper 69: Developing a Security Policy for the Use of CCTV in the Northern Border University

Abstract: The use of closed-circuit television (CCTV) in universities is a challenging task due to global strong opposition to its implementation at education institutions. The ministry of higher education of Kingdom of Saudi Arabia (KSA) has initiated a plan for monitoring educational institutes across the Kingdom. Therefore, this paper proposes a new framework for developing a comprehensive security policy for using CCTV in the Northern Border University, which streamlines the implementation, usage, and securing of the CCTV footage contents. In this regard, a new policy was developed combining the principles of activity theory, international standards, and design science methodology. It considered six key elements from both theoretical and practical perspectives, namely government rules, technical aspects, training, security requirements, users, and legal issues. Based on them, a standard 12-principal policy was developed; to help organizations easily implement and evaluate the developed policy and secure the contents, the principles were classified into three categories: performance, security, and policy management. The findings showed that the implementation of the policy developed in this study not only improved the security measures of the university, but also built trust among the stakeholders due to the high internal security and effective evaluation of the surveillance system.

Author 1: Ahmad Alshammari

Keywords: Closed circuit television; security policy; surveillance; educational institutes

Paper 70: Enhanced Gravitational Search Algorithm Based on Improved Convergence Strategy

Abstract: Gravitational search algorithm (GSA) is one of the metaheuristic algorithms that has been popularly implemented in solving various optimization problems. The algorithm could perform better in highly nonlinear and complex optimization problems. However, GSA has also been reported to have a weak local search ability and slow searching speed to achieve its convergence. This research proposes two new parameters in order to improve GSA’s convergence strategy by improving its exploration and exploitation capabilities. The parameters are the mass ratio and distance ratio parameters. The mass ratio parameter is related to the exploration strategy, while the distance ratio parameter is related to the exploitation strategy of the enhanced GSA (eGSA). These two parameters are expected to create a good balance between the exploration and the exploitation strategies in eGSA. There are seven benchmark functions that have been tested on eGSA. The results have shown that eGSA has been able to produce good performance in the minimization of fitness values and execution times, compared with two other GSA variants. The testing results have shown that the enhancements made to GSA have successfully improved the algorithm’s convergence strategy. The improved convergence has also been able to improve the algorithm’s solution quality and the processing time. It is expected that eGSA could be applied in many fields and solve various optimization problems efficiently.

Author 1: Norlina Mohd Sabri

Author 2: Ummu Fatihah Mohd Bahrin

Author 3: Mazidah Puteh

Keywords: Enhanced gravitational search algorithm; variant; improved convergence; exploration; exploitation

Paper 71: Proposed Secure Activity Diagram for Software Development

Abstract: Unified Modeling Language (UML) activity diagrams are derived from use case diagrams. It becomes essential to incorporate security features and maintain consistency in the diagrams during analysis phase of Software Development Life Cycle (SDLC). As part of current software development practices, software security must be a constant effort. The activity diagrams are used to model business process. The detailed analysis of activity diagram is done. The challenge lies in viewing the main activity diagram from attacker's perspective and providing defense mechanism to mitigate the attacks. This paper presents an extension of the activity diagram named SecUML3Activity to provide security with Object Constraint Language (OCL) constraints using Five Primary Security Input Validation Attributes (FPSIVA) parameters for input validation. It also proposed three security color code notations and stereotypes in activity diagrams. White color is used to represent activity diagram in normal state. Red color in dotted line is used to represent attack activity components. Blue color with double line is used to represent the defensive activity components. The defense mechanism algorithm against SQL Injection (SQLI) attack, Cross Site Scripting (XSS) attack, DoS/ DDoS attack, access validation attack is provided. The mapping of Secure 3-Use Case diagram with SecUML3Activity diagram is done through mathematical modeling.

Author 1: Madhuri N. Gedam

Author 2: Bandu B. Meshram

Keywords: Unified modeling language; activity diagram; object constraint language; SQL injection; use case diagram

Paper 72: An Efficient Vision-based Approach for Optimizing Energy Consumption in Internet of Things and Smart Homes

Abstract: One of the primary forces for digital transformation is how quickly the world is changing. Additionally, and at a dizzying pace, the world economy is being transformed by digital technology. The billions of daily online connections between individuals, organizations, devices, data, and processes that generate economic activity are known as the "digital economy." The Internet, mobile technology, and the Internet of Things (IoT) all contribute to hyper-interconnection, or the growing connectivity of people, organizations, and machines, which is the foundation of the digital economy. Simultaneously with these developments, the demand for energy is more than the supply, which leads to energy shortage. In order to keep pace with energy demand, new strategies are being developed. As a result of the emergence and expansion of smart homes, there is a growing need for digitization in applications such as energy efficient automation and safety. With the increase in the amount of electricity consumed and the introduction of new energy sources, the reduction of electricity costs for households becomes increasingly important. Basically, this article uses machine vision technology. In this paper, a YOlO method is used for facial recognition. And compared to all kinds of YOlO methods, the YOlOv5n method was the fastest and most efficient method. So, by using the YOlOv5s method on the Jetson Nano platform, it creates the possibility of authenticating the residents of the houses to identify them to turn on or off the sources of energy consumption in the houses. Therefore, the presented system is designed with the aim of optimizing energy consumption in houses and with the aim of ensuring the safety of the residents of the houses.

Author 1: LIU Chenguang

Keywords: IoT; Internet of things; digital economics; smart cities; digitization; machine vision; YOLO; YOLOv5n

Paper 73: Application of Medical Brain CT/MRI Image Fusion Algorithm based on Neural Network

Abstract: In recent years, fused images have been developed for fast processing of medical images, which provide a more reliable basis for reducing the burden on physicians because they can contain multiple times the image information. In order to achieve fast and accurate recognition results in medical image recognition, avoid similar blocks and shadow fitting in CT/MR fusion images, and improve the entire medical system, in this study, CT/MRI image fusion of brain images is studied based on algorithms generated by Convolutional Neural Network (CNN). The study utilizes Rolling Guidance Filter (RGF) to divide medical CT/MRI images into two parts, one of which is used for model training and the other for image fusion. In the experiments, the results of all three experiments are compared with the Nonsub Sampled Contourlet Transform - Piecewise Convolutional Neural Network (NSCT - PCNN), and the CNN-RGF MI/ IE/SSIM/AG values of CNN-RGF are superior compared to the conventional algorithm of NSCT-RCNN with an average improvement of 10.0% and above, and the resulting CNN-RGF observed meningitis, hydrocephalus, and cerebral infarction with an average of 24.8% higher compared to NSCT-RCNN. The outcomes show that for brain image fusion and detection, the CNN-RGF approach put forward in the study performs better.

Author 1: Dan Yang

Keywords: Convolutional neural network; image; integration; CT; MRI

Paper 74: Motion Path Planning of Wearable Lower Limb Exoskeleton Robot Based on Feature Description

Abstract: Wearable lower extremity exoskeleton robot is a kind of training equipment designed for the disabled or powerless in the lower extremity. In order to improve the environmental adaptability of the robot and better meet the use habits of patients, it is necessary to plan and design the movement path, and a movement path planning model of wearable lower extremity exoskeleton robot based on feature description is proposed, which describes the objects with different wearing frequencies and training intensities. Taking the wearer's natural walking gait as the constraint feature quantity and the control object model, the spatial planning and design of exoskeleton structures such as hip joint, knee joint and ankle joint are adopted, and the traditional single-degree-of-freedom rotating pair is replaced by a four-bar mechanism, which improves the bionic performance of the knee joint. Combining the feature description and the spatial planning algorithm model, an error compensation method based on iterative least square method is adopted to identify geometric parameters. The feature identification model of robot moving path planning is constructed, and the adaptive strong coupling tracking identification and path planning of robot moving path are realized through feature description and spatial distance error identification results. The simulation test results show that the cooperative positioning error is reduced and the torque error is compensated in real time by using this method to plan the movement path of the wearable lower limb exoskeleton robot, which makes the robot obtain better movement planning effect and enhance the stability of the mechanism.

Author 1: Ying Wang

Author 2: Songyu Sui

Keywords: Feature description; wearable lower limb exoskeleton robot; motion path planning; least square identification; geometric parameter

Paper 75: Robust Analysis of IT Infrastructure's Log Data with BERT Language Model

Abstract: Now-a-days, failure detection and prediction have become a significant research focus on enhancing the reliability and availability of IT infrastructure components. Log analysis is an emerging domain aimed at diminishing downtime caused by IT infrastructure components' failure. However, it can be challenging due to poor log quality and large data sizes. The proposed system automatically classifies logs based on log level and semantic analysis, allowing for a precise understanding of the meaning of log entries. Using the BERT pre-trained model, semantic vectors are generated for various IT infrastructures, such as Server Applications, Cloud Systems, Operating Systems, Supercomputers, and Mobile Systems. These vectors are then used to train machine learning (ML) classifiers for log categorization. The trained models are competent in classifying logs by comprehending the context of different types of logs. Additionally, semantic analysis outperforms sentiment analysis when dealing with unobserved log records. The proposed system significantly reduces engineers' day-to-day error-handling work by automating the log analysis process.

Author 1: Deepali Arun Bhanage

Author 2: Ambika Vishal Pawar

Keywords: System log; log analysis; BERT; classification; failure prediction; failure detection

Paper 76: Application of the Learning Set for the Detection of Jamming Attacks in 5G Mobile Networks

Abstract: Jamming attacks represent a significant problem in 5G mobile networks, requiring an effective detection mechanism to ensure network security. This study focused on finding effective methods for detecting these attacks using machine learning techniques. The effectiveness of Ensemble Learning and the XGBOOST-Ensemble Learning combination was evaluated by comparing their performance to other existing approaches. To carry out this study, the WSN-DS database, widely used in attack detection, was used. The results obtained show that the hybrid method, XGBOOST-Ensemble Learning, outperforms other approaches, including those described in the literature, with an accuracy ranging from 99.46% to 99.72%. This underlines the effectiveness of this method for accurately detecting jamming attacks in 5G networks. By using advanced machine learning techniques, the present study helps strengthen the security of 5G mobile networks by providing a reliable mechanism to detect and prevent jamming attacks. These encouraging results also open avenues for future research to further improve the accuracy and effectiveness of attack detection in radiocommunication in general and specifically in 5G networks, thereby ensuring better protection for next-generation wireless communications.

Author 1: Brou Médard KOUASSI

Author 2: Vincent MONSAN

Author 3: Abou Bakary BALLO

Author 4: Kacoutchy Jean AYIKPA

Author 5: Diarra MAMADOU

Author 6: Kablan Jérome ADOU

Keywords: Jamming attacks; 5G mobile networks; ensemble learning; XGBOOST-ensemble learning; attack detection

Paper 77: Hybrid Global Structure Model for Unraveling Influential Nodes in Complex Networks

Abstract: In graph analytics, the identification of influential nodes in real-world networks plays a crucial role in understanding network dynamics and enabling various applications. However, traditional centrality metrics often fall short in capturing the interplay between local and global network information. To address this limitation, the Global Structure Model (GSM) and its improved version (IGSM) have been proposed. Nonetheless, these models still lack an adequate representation of path length. This research aims to enhance existing approaches by developing a hybrid model called H-GSM. The H-GSM algorithm integrates the GSM framework with local and global centrality measurements, specifically Degree Centrality (DC) and K-Shell Centrality (KS). By incorporating these additional measures, the H-GSM model strives to improve the accuracy of identifying influential nodes in complex networks. To evaluate the effectiveness of the H-GSM model, real-world datasets are employed, and comparative analyses are conducted against existing techniques. The results demonstrate that the H-GSM model outperforms these techniques, showcasing its enhanced performance in identifying influential nodes. As future research directions, it is proposed to explore different combinations of index styles and centrality measures within the H-GSM framework.

Author 1: Mohd Fariduddin Mukhtar

Author 2: Zuraida Abal Abas

Author 3: Amir Hamzah Abdul Rasib

Author 4: Siti Haryanti Hairol Anuar

Author 5: Nurul Hafizah Mohd Zaki

Author 6: Ahmad Fadzli Nizam Abdul Rahman

Author 7: Zaheera Zainal Abidin

Author 8: Abdul Samad Shibghatullah

Keywords: Centrality indices; combination; hybrid; global structure model; influential nodes

Paper 78: Multi-Granularity Tooth Analysis via Faster Region-Convolutional Neural Networks for Effective Tooth Detection and Classification

Abstract: In image classification, multi-granularity refers to the ability to classify images with different levels of detail or resolution. This is a challenging task because the distinction between subcategories is often minimal, needing a high level of visual detail and precise representation of the features specific to each class. In dental informatics, and more specifically tooth classification poses many challenges due to overlapping teeth, varying sizes, shapes, and illumination levels. To address these issues, this paper considers various data granularity levels since a deeper level of details can be acquired with increased granularity. Three tooth granularity levels are considered in this study named Two Classes Granularity Level (2CGL), Four Classes Granularity Level (4CGL), and Seven Classes Granularity Level (7CGL) to analyze the performance of teeth detection and classification at multi-granularity levels in Granular Intra-Oral Image (GIOI) dataset. Subsequently, a Faster Region-Convolutional Neural Network (FR-CNN) based on three ResNet models is proposed for teeth detection and classification at multi-granularity levels from the GIOI dataset. The FR-CNN-ResNet models exploit the effect of the tooth classification granularity technique to empower the models with accurate features that lead to improved model performance. The results indicate a remarkable detection effect in investigating the granularity effect on the FR-CNN-ResNet model's performance. The FR-CNN-ResNet-50 model achieved 0.94 mAP for 2CGL, 0.74 mAP for 4CGL, and 0.69 mAP for 7CGL, respectively. The findings demonstrated that multi-granularity enables flexible and nuanced analysis of visual data, which can be useful in a wide range of applications.

Author 1: Samah AbuSalim

Author 2: Nordin Zakaria

Author 3: Salama A Mostafa

Author 4: Yew Kwang Hooi

Author 5: Norehan Mokhtar

Author 6: Said Jadid Abdulkadir

Keywords: Dental informatics; intra-oral image; deep learning; faster region-convolutional neural network; classification; granularity level; tooth detection

Paper 79: A Hybrid Approach for Underwater Image Enhancement using CNN and GAN

Abstract: Underwater image-capturing technology has advanced over the years, and varieties of artificial intelligence-based applications have been developed on digital and synthetic images. The low-quality and low-resolution underwater images are challenging factors for use in existing image processing in computer vision applications. Degraded or low-quality photos are common issues in the underwater imaging process due to natural factors like low illumination and scattering. The recent techniques use deep learning architectures like CNN, GAN, or other models for image enhancement. Although adversarial-based architectures provide good perceptual quality, they performed worse in quantitative tests compared with convolutional-based networks. A hybrid technique is proposed in this paper that blends both designs to gain advantages of the CNN and GAN architectures. The generator component produces or makes images, which contributes to the creation of a sizable training set. The EUVP dataset is used for experimentation for model training and testing. The PSNR score was observed to measure the visual quality of the resultant images produced by models. The proposed system was able to provide an improved image with a higher PSNR score and SSIM score with state-of-the-art methods.

Author 1: Aparna Menon

Author 2: R Aarthi

Keywords: Convolutional neural network (CNN); generative adversarial networks (GAN); enhancing underwater visual perception (EUVP); underwater images; image enhancement; computer vision; artificial intelligence

Paper 80: End-to-End Real-time Architecture for Fraud Detection in Online Digital Transactions

Abstract: The banking sector is witnessing a fierce concurrence characterized by changing business models, new entrants such as FinTechs, and new customer behaviors. Financial institutions try to adapt to this trend and invent new ways and channels to reach and interact with their customers. While banks are opening their services to avoid missing this shift, they become naturally exposed to fraud attempts through their digital banking platforms. Therefore, fraud prevention and detection are considered must-have capabilities. Detecting fraud at an optimal time requires developing and deploying scalable learning systems capable of ingesting and analyzing vast volumes of streaming records. Current improvements in data analytics algorithms and the advent of open-source technologies for big data processing and storage bring up novel avenues for fraud identification. In this article, we provide a real-time architecture for detecting transactional fraud via behavioral analysis that incorporates big data analysis techniques such as Spark, Kafka, and h2o with an unsupervised machine learning (ML) algorithm named Isolation Forest. The results of experiments on a significant dataset of digital transactions indicate that this architecture is robust, effective, and reliable across a large set of transactions yielding 99% of accuracy, and a precision of 87%.

Author 1: ABBASSI Hanae

Author 2: BERKAOUI Abdellah

Author 3: ELMENDILI Saida

Author 4: GAHI Youssef

Keywords: Online fraud; big data analytics; fraud detection; behavior analysis; isolation forest

Paper 81: Prediction of Anti-inflammatory Activity of Bio Copper Nanoparticle using an Innovative Soft Computing Methodology

Abstract: The objective of this work is to use a novel soft computing approach to predict the anti-inflammatory effect of bio copper nanoparticles. Using a modified technique, various doses of the Musa sapientum extract and copper nanoparticles were examined for their anti-inflammatory capabilities. Protein denaturation was evaluated, and an inhibition percentage was computed. The outcomes demonstrated that the quantity of copper nanoparticles raised the inhibition percentage, indicating a greater anti-inflammatory efficacy. In order to forecast the anti-inflammatory action based on the input variables of contact duration, operating temperature, and beginning concentration, an artificial neural network (ANN) was created. Using experimental data, the ANN model was developed, tested, and its performance assessed. The outcomes showed that the ANN model has a high degree of accuracy in predicting the anti-inflammatory action. In the context of summary, copper nanoparticles produced by Musa sapientum show considerable anti-inflammatory action. The ANN model and the suggested soft computing technique, which included the creation of copper nanoparticles, made an accurate prediction of the anti-inflammatory capabilities. This study aids in creating new methods for estimating the efficacy of bioactive nanoparticles in diverse therapeutic uses, such as the treatment of inflammation.

Author 1: Dyuti Banerjee

Author 2: G. Kiran Kumar

Author 3: Farrukh Sobia

Author 4: Subuhi Kashif Ansari

Author 5: Anuradha. S

Author 6: R. Manikandan

Keywords: Copper; nanoparticles; green synthesis; prediction; artificial neural network

Paper 82: Automatic Essay Scoring for Arabic Short Answer Questions using Text Mining Techniques

Abstract: Automated Essay Scoring (AES) systems involve using a specially designed computing program to mark students’ essays. It is a form of online assessment supported by natural language processing (NLP). These systems seek to exploit advanced technologies to reduce the time and effort spent on the exam scoring process. These systems have been applied in several languages, including Arabic. Nevertheless, the applicable NLP techniques in Arabic AES are still limited, and further investigation is needed to make NLP suitable for Arabic to achieve human-like scoring accuracy. Therefore, this comparative empirical experimental study tested two word-embedding deep learning approaches, namely BERT and Word2vec, along with a knowledge-based similarity approach; Arabic WordNet. The study used the Cosine similarity measure to provide optimal student answer scores. Several experiments were conducted for each of the proposed approaches on two available Arabic short answer question datasets to explore the effect of the stemming level. The quantitative results of this study indicated that advanced models of contextual embedding can improve the efficiency of Arabic AES as the meaning of words can differ in the different contexts. Therefore, serve as a catalyst for future research based on contextual embedding models, as the BERT approach achieved the best Pearson Correlation (.84) and RMSE (1.003). However, this research area needs further investigation to increase the accuracy of Arabic AES to become a practical online scoring system.

Author 1: Maram Meccawy

Author 2: Afnan Ali Bayazed

Author 3: Bashayer Al-Abdullah

Author 4: Hind Algamdi

Keywords: Arabic language; Automated Essay Scoring (AES); Automated Scoring (AS); Educational Technologies; NLP

Paper 83: Combination of Adaptive Neuro Fuzzy Inference System and Machine Learning Algorithm for Recognition of Human Facial Expressions

Abstract: A face recognition system's initial three processes are face detection, feature extraction, and facial expression recognition. The initial step of face detection involves colour model skin colour detection, lighting adjustment to achieve uniformity on the face, and morphological techniques to maintain the necessary face region. To extract facial characteristics such the eyes, nose, and mouth, the output of the first step is employed. Third-step methodology using automated face emotion recognition. This study's goal is to apply the Adaptive Neuro Fuzzy Inference System (ANFIS) algorithm to increase the precision of the current face recognition systems. For the purpose of removing noise and unwanted information from the data sets, independent data sets and a pre-processing technique are built in this study based on color, texture, and shape, to determine the features of the face. The output of the three-feature extraction process is given to the ANFIS model as input. By using our training picture data sets, it has already been trained. This model receives a test image as input, then evaluates the three aspects of the input image, and then recognizes the test image based on correlation. The determination of whether input has been authenticated or not is made using fuzzy logic. The proposed ANFIS method is compared to the existing methods such as Minimum Distance Classifier (MDC), Support Vector Machine (SVM), Case Based Reasoning (CBR) with the following quality measure like error rate, accuracy, precision, recall. Finally, the performance is analyzed by combining all feature extractions with existing classification methods such as MDC, KNN (K-Nearest Neighbour), SVM, ANFIS and CBR. Based on the performance of classification techniques, it is observed that the face detection failures are reduced, such that overall accuracy for CBR is 92% and it is 97% in ANFIS.

Author 1: B. Dhanalaxmi

Author 2: B. Madhuravani

Author 3: Yeligeti Raju

Author 4: C. Balaswamy

Author 5: A. Athiraja

Author 6: G. Charles Babu

Author 7: T. Samraj Lawrence

Keywords: ANFIS; Image processing; face recognition; feature extraction; fuzzy logic

Paper 84: Comparative Analysis of DIDIM and IV Approaches using Double Least Squares Method

Abstract: Usually, identifying dynamic parameters for robots involves utilizing the Inverse Dynamic Model (IDM) which is linear in relation to the parameters being identified, alongside Linear Least Squares (LLS) methods. To implement this approach, precise measurements of both torque and position must be obtained at a high frequency. Additionally, velocities and accelerations must be estimated by implementing a band-pass filtering technique on the position data. Given the presence of noise in the observation matrix and the closed-loop nature of the identification process, we have modified the Instrumental Variable (IV) method to address the issue of noisy observations. A novel identification technique, named (Direct and Inverse Dynamic Identification Model) DIDIM, which requires only torque measurements as input variables, has recently been successfully applied to a 6-degree-of-freedom industrial robot. DIDIM employs a closed-loop output error approach that utilizes closed-loop simulations of the robot. The experimental results reveal that the IV and DIDIM methods exhibit numerical equivalence. In this paper, we conduct a comparison of these two methods using a double step least squares (2SLS) analysis. We experimentally validate this study using a 2-degree-of-freedom planar robot.

Author 1: Fadwa SAADA

Author 2: David DELOUCHE

Author 3: Karim CHABIR

Author 4: Mohamed Naceur ABDELKRIM

Keywords: Identification; double least squares; instrumental variable; DIDIM method; robotics dynamics

Paper 85: Skin Cancer Classification using Delaunay Triangulation and Graph Convolutional Network

Abstract: Oftentimes, many people or even medical workers misdiagnose skin cancer, which may lead to malpractice and thus, resulting in delayed recovery or life-threatening complications. In this research, a Graph Convolutional Network (GCN) method is proposed as a classification model and Delaunay triangulation as its feature extraction method to classify various types of skin cancers. Delaunay triangulation serves the purpose of boundary extraction, and this implementation allows the model to focus only on the cancerous lesion and ignore the skin around it. This way, the types of skin cancer can be predicted more accurately. Furthermore, GCN offers many advantages in medical image analysis over traditional CNN models. GCN can model interactions between different regions and structures in an image and perform messaging between nodes, whereas CNN is not explicitly designed to do such thing. Other than that, GCN can also leverage transfer learning and few-shot learning techniques to address the challenges of limited annotated medical image datasets. However, the result shows that the proposed model tends to overfit and is unable to generate correct predictions for new skin cancer images. There are several reasons that could lead the model to overfit, such as imbalance data, incorrect feature extraction, insufficient features for data prediction, or the data containing noise.

Author 1: Caroline Angelina Sunarya

Author 2: Jocelyn Verna Siswanto

Author 3: Grace Shirley Cam

Author 4: Felix Indra Kurniadi

Keywords: Skin cancer; Delaunay triangulation; graph convolutional network; GCN; multilabel image classification; convolutional neural network; CNN

Paper 86: Fertigation Technology Meets Online Market: A Multipurpose Mobile App for Urban Farming

Abstract: In a world where smartphones dominate the market and provide opportunities to a vast population, this work introduces an innovative application that enables users to order irrigated vegetable crops from urban farmers. The application utilizes simple fertigation system technology, which can be easily implemented in small areas such as homes. Currently, consumers and sellers rely on traditional methods like paper or other means to place orders for vertical farming. However, research shows that these methods are unreliable and only offer temporary relevance. Additionally, the traditional agriculture supply chain diminishes the appeal of urban farming as its benefits do not outweigh the disadvantages. The primary objective of this application is to promote the concept of urban farming by creating an online marketplace that bypasses traditional methods. This allows consumers to directly order from the farmers themselves, serving as an alternative to the agricultural supply chain. Furthermore, a monitoring system has been integrated into the application as an additional tool, enabling farmers to remotely monitor and control their farms. This feature is particularly beneficial for urban farmers with farms in multiple locations who may lack the time to physically visit each one.

Author 1: Jamil Abedalrahim Jamil Alsayaydeh

Author 2: Mohd Faizal bin Yusof

Author 3: Asyraf Salmi

Author 4: Adam Wong Yoon Khang

Author 5: Safarudin Gazali Herawan

Keywords: Vertical farming; mobile application; online market; farm monitoring; urban farmer; agriculture supply chain

Paper 87: Offensive Language Identification in Low Resource Languages using Bidirectional Long-Short-Term Memory Network

Abstract: Offensive language identification is a critical task in today's digital era, enabling the development of effective content moderation systems. However, it poses unique challenges in low resource languages where limited annotated data is available. This research paper focuses on addressing the problem of offensive language identification specifically in the context of a low resource language, namely the Kazakh language. To tackle this challenge, we propose a novel approach based on Bidirectional Long-Short-Term Memory (BiLSTM) networks, which have demonstrated strong performance in natural language processing tasks. By leveraging the bidirectional nature of the BiLSTM architecture, we capture both contextual dependencies and long-term dependencies in the input text, enabling more accurate offensive language identification. Our approach further utilizes transfer learning techniques to mitigate the scarcity of annotated data in the low resource setting. Through extensive experiments on a Kazakh offensive language dataset, we demonstrate the effectiveness of our proposed approach, achieving state-of-the-art results in offensive language identification in the low resource Kazakh language. Moreover, we analyze the impact of different model configurations and training strategies on the performance of our approach. The findings from our study provide valuable insights into offensive language identification techniques in low resource languages and pave the way for more robust content moderation systems tailored to specific linguistic contexts.

Author 1: Aigerim Toktarova

Author 2: Aktore Abushakhma

Author 3: Elvira Adylbekova

Author 4: Ainur Manapova

Author 5: Bolganay Kaldarova

Author 6: Yerzhan Atayev

Author 7: Bakhyt Kassenova

Author 8: Ainash Aidarkhanova

Keywords: Offensive language; natural language processing; low resource language; machine learning; deep learning; classification

Paper 88: Query-Focused Multi-document Summarization Survey

Abstract: With the exponential growth of textual information on the web and in multimedia, query-focused multi-document summarization (QFMS) has emerged as a critical research area. QFMS aims to generate concise summaries that address user queries and satisfy their information needs. This paper provides a comprehensive survey of state-of-the-art approaches in QFMS, focusing specifically on graph-based and clustering-based methods. Each approach is examined in detail, highlighting its advantages and disadvantages. The survey covers ranking algorithms, sentence selection techniques, redundancy removal methods, evaluation metrics, and available datasets. The principal aim of this paper is to present a thorough analysis of QFMS approaches, providing researchers and practitioners with valuable insights into the field. By surveying existing techniques, the paper identifies the challenges and issues faced in QFMS and discusses potential future directions. Moreover, the paper emphasizes the importance of addressing coherency, ambiguity, vague references, evaluation methods, redundancy, and diversity in QFMS. Performance standards and competing approaches are also discussed, showcasing the advancements made in QFMS. The paper acknowledges the need for improving summarization coherence, readability, and semantic efficiency, while balancing compression ratios and summarizing quality. Additionally, it highlights the potential of hybrid methods and the integration of extractive and abstractive techniques to achieve more human-like summaries.

Author 1: Entesar Alanzi

Author 2: Safa Alballaa

Keywords: Text summarization; query-based extractive text summarization; multi-document; graph-based approach; clustering-based approach

Paper 89: Ensuring Information Security of Web Resources Based on Blockchain Technologies

Abstract: This project examines how blockchain technology can enhance data security and reliability for web applications. In this article, ways to improve data security on online course platforms that utilize blockchain technology are explored. To clarify, online course platforms are web-based applications that enable users to access course materials online. These platforms often deal with sensitive data which, if compromised, can cause significant harm to users. Unfortunately, this information is often the target of fraudulent operations and illegal actions aimed at stealing personal data that can be used for authentication on various platforms. This article identifies the weaknesses of these sites and discusses the importance of using complex technologies to safeguard web resources effectively. This research explores how blockchain technology can protect from common web application attacks, which are often aimed at the user authorization process involving the transmission of identification and authentication data from the user to the website database. The study outlines the key components of blockchain technology, including hash function, hash value, data structure, and blockchain classification. Additionally, the study presents a transaction block model for a web course developed using blockchain technology.

Author 1: Barakova Aliya

Author 2: Ussatova Olga

Author 3: Begimbayeva Yenlik

Author 4: Ibrahim Sogukpinar

Keywords: Information security; data security; website protection; blockchain; network attacks; hash functions; web applications

Paper 90: A Hybrid Multiple Indefinite Kernel Learning Framework for Disease Classification from Gene Expression Data

Abstract: In recent years, Machine Learning (ML) techniques have been used by several researchers to classify diseases using gene expression data. Disease categorization using heterogeneous gene expression data is often used for defining critical problems such as cancer analysis. A variety of evaluated factors known as genes are used to characterize the gene expression data gathered from DNA microarrays. Accurate classification of genetic data is essential to provide accurate treatments to sick people. A large number of genes can be viewed simultaneously from the collected data. However, processing this data has some limitations due to noises, redundant data, frequent errors, increased complexity, smaller samples with high dimensionality, difficult interpretation, etc. A model must be able to distinguish the features in such heterogeneous data with high accuracy to make accurate predictions. So this paper presents an innovative model to overcome these issues. The proposed model includes an effective multiple indefinite kernel learning based model for analyze the gene expression microarray data, then an optimized kernel principal component analysis (OKPCA) to select best features and hybrid flow-directed arithmetic support vector machine (SVM)-based multiple infinite kernel learning (FDASVM-MIKL) model for classification. Flow direction and arithmetic optimization algorithms are combined with SVM to increase classification accuracy. The proposed technique has an accuracy of 99.95%, 99.63%, 99.60%, 99.51%, and 99.79% using the datasets including colon, Isolet, ALLAML, Lung_cancer, and Snp2 graph.

Author 1: Swetha S

Author 2: Srinivasan G N

Author 3: Dayananda P

Keywords: Gene expression; optimized kernel principle component analysis; multiple indefinite kernel learning; flow direction algorithm based support vector machine; arithmetic optimization algorithm

Paper 91: A 3D Processing Technique to Detect Lung Tumor

Abstract: In this paper, the authors introduce a new segmentation technique based on U-NET algorithm from the deep learning used for lung cancer segmentation, which is the main challenge that medical Staff confront in their diagnosis process. The goal is to develop an ideal segmentation that enables medical personnel to distinguish the various tumor components using the completely U-NET convolution network architecture, which is the most effective. First, the regions of interest (ROI) in the 2D slides are established by an expert using the syngovia application of the Siemens. In this pre-processing step, the cancer area is isolated from its surroundings, and is used as a training model for U-NET algorithm. Second, the 2D U-NET model is used to segment the DICOM images (Digital Imaging and Communications in Medicine) into homogeneous regions. Finally, the post processing step has been used to obtain the 3D CT scan (computerized tomography) from the 2D slices. The segmentation results from the proposed method applied on biomedical images from nuclear medicine and radiotherapy that are extracted from the archiving system of the Institute of Salah Azaiez from Tunisia. The segmentation results are validated, and the prediction accuracy for the available test data is evaluated. Finally, a comparison study with other existing techniques is presented. The experimental results demonstrate the superiority of the used U-NET architecture applied either for 2D or for 3D image segmentation.

Author 1: Nabila ELLOUMI

Author 2: Slim Ben CHAABANE

Author 3: Hassan SEDDIK

Author 4: TOUNSI Nadra

Keywords: Deep learning U-NET architecture; 3D CT scan (computerized tomography); DICOM images (Digital Imaging and Communications in Medicine); 2D slices; ROI (regions of interest)

Paper 92: Prediction of Breast Cancer using Traditional and Ensemble Technique: A Machine Learning Approach

Abstract: Breast cancer is a prevalent and potentially life-threatening disease that affects millions of individuals worldwide. Early detection plays a crucial role in improving patient outcomes and increasing the chances of survival. In recent years, machine learning (ML) techniques have gained significant attention in the field of breast cancer detection and diagnosis due to their ability to analyze large and complex datasets, extract meaningful patterns, and facilitate accurate classification. This research focuses on leveraging ML algorithms and models to enhance breast cancer detection and provide more reliable diagnostic results in the real world. Two datasets from Kaggle have been used in this study and Decision tree (DT), Random Forest (RF), Logistic Regression (LR), K-Nearest Classifier (KNN) etc. are applied to identify potential breast cancer cases. On the first dataset, A, the test's accuracy using Logistic Regression, SVM, and Grid SearchCV was 95.614%, however in dataset B, the accuracy of Logistic Regression and Decision Tree increased to 99.270%. The accuracy of Boosting Decision Tree was 99.270% when compared to other algorithms. To defend the performances, various ensemble models are used. To assign the optimal parameters to each classifier, a hyper-parameter tweaking method is used. The experimental study examined the findings of recent studies and discovered that LRBO performed best, with the highest level of accuracy for predicting breast cancer being 95.614%.

Author 1: Tamanna Islam

Author 2: Amatul Bushra Akhi

Author 3: Farzana Akter

Author 4: Md. Najmul Hasan

Author 5: Munira Akter Lata

Keywords: Breast cancer; prediction; machine learning algorithms; ensemble models; voting; stacking

Paper 93: Research on Settlement Prediction of Building Foundation in Smart City Based on BP Network

Abstract: In the construction process of high-rise buildings, it is necessary to predict the settlement and deformation of the foundation, and the current prediction methods are mainly based on empirical theoretical calculations and methods and more accurate numerical analysis methods. In the face of the interference of complex and ever-changing terrain and parameter values on prediction methods, in order to accurately determine the settlement of building foundations, this study designed a smart city building foundation settlement prediction method based on BP neural network. Firstly, a real-time dynamic monitoring unit for building foundation settlement was constructed using Wireless Sensor Network (WSN) technology. Then, the monitoring data was used to calculate the relevant parameters of building foundation settlement through layer sum method. Finally, input the monitoring data into the BP network results, adjust the weights of the output layer and hidden layer using settlement related parameters, and output the settlement prediction results of the smart city building foundation through training. The study selected average error and prediction time as evaluation criteria to test the feasibility of the method proposed in this article. This method can effectively predict foundation settlement, with an average prediction error always less than 4% and a prediction process time always less than 49ms.

Author 1: Luyao Wei

Keywords: Smart city; intelligent architecture; foundation settlement; settlement prediction; BP neural network; parameter

Paper 94: A Classified Warning Method for Heavy Overload in Distribution Networks Considering the Characteristics of Unbalanced Datasets

Abstract: In order to achieve heavy overload warning and capacity planning for the distribution network, it is necessary to classify the heavy overload warning of the distribution network. A distribution network with heavy overload classification warning method based on imbalanced dataset feature extraction is proposed. Screening the feature indicator set related to distribution network overload, constructing a hierarchical prediction framework for distribution network load situation, combining information such as power distribution points, road construction, municipal planning, and power load distribution to form distribution network capacity planning and line renovation plans. Based on K-means clustering, the undersampling method is used to extract features from the unbalanced dataset of distribution network overload classification, using decision trees as the basic learning unit. It includes multiple decision trees trained by Bagging integrated learning theory and random subspace method. The random forest algorithm is used to realize the feature detection and distribution network capacity planning of distribution network weight overload grading, and the grading early warning of distribution network weight overload is realized according to the capacity planning results. Tests have shown that this method has good accuracy in predicting electrical loads and can effectively solve the problem of excess capacity caused by light or no load, improving the ability of heavy overload warning and capacity planning in the distribution network.

Author 1: Guohui Ren

Keywords: Imbalanced data; feature extraction; distribution network; overload classification warning

Paper 95: A Novel Method for Myocardial Image Classification using Data Augmentation

Abstract: Myocarditis is an important public health concern since it can cause heart failure and abrupt death. It can be diagnosed with magnetic resonance imaging (MRI) of the heart, a non-invasive imaging technology with the potential for operator bias. The study provides a deep learning-based model for myocarditis detection using CMR images to support medical professionals. The proposed architecture comprises a convolutional neural network (CNN), a fully-connected decision layer, a generative adversarial network (GAN)-based algorithm for data augmentation, an enhanced DE for pre-training weights, and a reinforcement learning-based method for training. We present a new method of employing produced images for data augmentation based on GAN to improve the classification performance of the provided CNN. Unbalanced data is one of the most significant classification issues, as negative samples are more than positive, decimating system performance. To solve this issue, we offer an RL-based training method that learns minority class examples with attention. In addition, we tackle the challenges associated with the training step, which typically relies on gradient-based techniques for the learning process; however, these methods often face issues like sensitivity to initialization. To start the BP process, we present an improved differential evolution (DE) technique that leverages a clustering-based mutation operator. It recognizes a successful cluster for DE and applies an original updating strategy to produce potential solutions. We assess our suggested model on the Z-Alizadeh Sani myocarditis dataset and show that it outperforms other methods.

Author 1: Qing kun Zhu

Keywords: Myocarditis; generative adversarial network; data augmentation; differential evolution

Paper 96: Design and Application of Online Courses under the Threshold of Smart Innovation Education

Abstract: With the rapid development of the Internet and the growing demand for education, a new online teaching mode, massive open online courses (MOOC), emerged in 2012. To address the problems of sparse data and poor recommendation effect in online course recommendation, this paper introduces deep learning into course recommendation and proposes an auxiliary information-based neural network model (IUNeu), on the basis of which a collaborative neural network filtering model (FIUNeu) is obtained by improving it. Firstly, the principles and technical details of the deep learning base model are studied in depth to provide technical support for course recommendation models and online learning recommendation systems. In this paper, based on the existing neural matrix decomposition model (NeuMF), we combine user information and course information and consider the interaction relationship between them to improve the accuracy of the model to represent users and courses. The neural network model of auxiliary information (IUNeu) is incorporated into the online learning platform, and the system development is completed with the design of front and back-end separation, realizing the functions of the online learning module, course collection module, course recommendation module, and resource download module. Finally, the experimental results are analyzed: under the same experimental conditions, the test experiments are repeated 10 times, and the RMSE calculation results are averaged. The RMSE value of the neural network collaborative filtering model (FIUNeu) proposed in this paper based on deep learning is 0.85517, which is the best performance and has a high accuracy rate of rating prediction, and is useful for alleviating the data sparsity problem.

Author 1: Qin Wang

Author 2: Anya Xiong

Author 3: Huirong Zhu

Keywords: Massive open online courses (MOOC); deep learning; collaborative neural network filtering model (FIONeu); course recommendation; online learning recommendation system

Paper 97: Object Detection-based Automatic Waste Segregation using Robotic Arm

Abstract: Today's overpopulation and fast urbanization present a significant challenge for developing countries in the form of excessive garbage generation. Managing waste is essential in creating sustainable and habitable communities, but it remains an issue for developing countries. Finding an efficient smart waste management system is a challenge in current research. In recent years, robots and artificial intelligence have influenced a wide range of industries, especially waste management. This research proposes a waste segregation system that integrates the robot arm and YOLOv6 object detection model to automatically sort the garbage according to its type and achieve real-time requirements. The proposed algorithm utilizes the pros of the hardware-friendly architecture of YOLOv6 while keeping high detection accuracy in detecting and classifying garbage. Moreover, the proposed system creates a 3D model of a 4 DOF robotic arm by CAD tools. A new approach based on a geometric method is proposed to solve the inverse kinematics problem in order to precisely calculate the proper angles of the robot arm's joints via a unique solution with less computational time. The proposed system is evaluated on a modified TrashNet dataset with seven garbage classes. The experiments reveal that the proposed algorithm outperforms the other recent YOLO models in terms of precision, recall, F1 score, and model size. Furthermore, the proposed algorithm consumes approximately fractions of a second for picking up and placing a single object in its proper basket.

Author 1: Azza Elsayed Ibrahim

Author 2: Rasha Shoitan

Author 3: Mona M. Moussa

Author 4: Heba A. Elnemr

Author 5: Young Im Cho

Author 6: Mohamed S. Abdallah

Keywords: Smart recycling; inverse kinematics; object detection; 4 DOF robotic arm; YOLOV6

Paper 98: Diversity-based Test Case Prioritization Technique to Improve Faults Detection Rate

Abstract: Regression testing is an important task in software development, but it is often associated with high costs and increased project expenses. To address this challenge, prioritizing test cases during test execution is essential as it aims to swiftly identify the hidden faults in the software. In the literature, several techniques for test case prioritization (TCP) have been proposed and evaluated. However, existing weight-based TCP techniques often overlook the true diversity coverage of test cases, resulting in the use of average-based weighting practices and a lack of systematic calculation for test case weights. Our research revolves around prioritizing test cases by considering multiple code coverage criteria. The study presents a novel diversity technique that calculates a diversity coverage score for each test case. This score serves as a weight to effectively rank the test cases. To evaluate the proposed technique, an experiment was conducted using five open-source programs and measured its performance in terms of the average percentage of fault detection (APFD). A comparison was made against an existing technique. The results revealed that the proposed technique significantly improved the fault detection rate compared to the existing approach. It is worth noting that this study is the first of its kind to incorporate the true diversity score of test cases into the TCP process. The findings of our research make valuable contributions to the field of regression testing by enhancing the effectiveness of the testing process through the utilization of diversity-based weighting techniques.

Author 1: Jamal Abdullahi Nuh

Author 2: Tieng Wei Koh

Author 3: Salmi Baharom

Author 4: Mohd Hafeez Osman

Author 5: Lawal Babangida

Author 6: Sukumar Letchmunan

Author 7: Si Na Kew

Keywords: Regression testing; fault detection; test case prioritization; test case diversity; test case coverage; species diversity

Paper 99: Exploring the Impact of Hybrid Recommender Systems on Personalized Mental Health Recommendations

Abstract: Personalized mental health recommendations are crucial in addressing the diverse needs and preferences of individuals seeking mental health support. This research aims to study the investigates the impact of hybrid recommender systems on the provision of personalized recommendations for mental health interventions. This paper explores the integration of various recommendation techniques, including collaborative filtering, content-based filtering, and knowledge-based filtering, within the hybrid system to leverage their respective strengths for Personalized Mental Health Recommendations. Additionally, this paper discusses the challenges and considerations involved in combining multiple techniques, such as data integration and algorithm selection for Hybrid Recommender System for this domain. Furthermore, this paper also discusses the data sources that are typically used in hybrid recommender systems for mental health and evaluation metrics that are employed to assess the effectiveness of the hybrid recommender system. Future research opportunities, including incorporating emerging technologies and leveraging novel data sources, are identified to further enhance the performance and relevance of hybrid recommender systems in the mental health domain. The findings of this research contribute to the advancement of personalized mental health support and the development of effective recommendation systems tailored to individual mental health needs.

Author 1: Idayati Mazlan

Author 2: Noraswaliza Abdullah

Author 3: Norashikin Ahmad

Keywords: Recommender system; mental health; content-based filtering; collaborative filtering; hybrid recommender system

Paper 100: Classification of Garlic Land Based on Growth Phase using Convolutional Neural Network

Abstract: Indonesian Government needs to monitor the realization of garlic land with production plans in several production areas at growth season. A previous study, which used Sentinel-1A satellite imagery and Convolutional Neural Networks to classify garlic land, needed more information on growth phases. The study aims to address that limitation by creating a garlic land classification model based on the growth phase using Convolutional Neural Networks. The dataset comprises 446 preprocessed Sentinel-2 images cross-referenced with drone ground truth data. The model used both VGG16 and VGG19 architectures. Hyperparameter tuning was applied to obtain optimal values. After evaluating three scenarios (VGG16 base model, modified VGG16, and modified VGG19), the best model was obtained from the modified VGG19, which had an accuracy rate of 81.81% and a loss function of 0.71. The study successfully classified garlic land based on growth phase, with a precision rate of 0.43 for initial growth and vegetation classes, and 0.22 for the harvest class. The study offers an alternative to monitoring garlic production throughout growth phases with satellite imagery and deep learning.

Author 1: Durrotul Mukhibah

Author 2: Imas Sukaesih Sitanggang

Author 3: Annisa

Keywords: Convolutional neural network; garlic; growth phase; horticulture; land classification; Sentinel-2; VGG

Paper 101: Evaluating Game Application Interfaces for Older Adults with Mild Cognitive Impairment

Abstract: A digital game is software that is used as alternative entertainment for older adults for brain training. In this study, a digital game prototype for older adults with mild cognitive impairment has been developed called EmoGame and illustrated. The game is intended to assist older adults who experience emotional and cognitive impairment that implement reminiscence therapy in the design of the user interface. Applications for older adults have been developed in many studies, but applications using a reminiscence therapy approach still need to be improved. User interface testing was carried out using the system usability scale (SUS). Interface testing with the SUS instrument was carried out in an organized and precisely measured using ten (10) questions as a benchmark for evaluation among twenty (20) respondents of, older adults. The results of the evaluation of the EmoGame prototype show an assessment score of 82, representing an excellent rating. Future work will improve the prototype to improve the design based on user feedback and iteratively improve the functionalities and interfaces and conduct a longitudinal study to investigate the effect of the games towards improving cognitive among older adults with mild cognitive impairment.

Author 1: Nita Rosa Damayanti

Author 2: Nazlena Mohamad Ali

Keywords: System usability scale; older adults; mild cognitive impairment

Paper 102: Intelligent Recommendation of Open Educational Resources: Building a Recommendation Model Based on Deep Neural Networks

Abstract: Information overload is a challenge for the development of online education. To address the problem of intelligent recommendation of educational resources, the study proposes an intelligent recommendation model of educational resources based on deep neural networks. First, a deep neural network-based custom recommendation model for educational resources is constructed after a multilayer perceptron-based prediction model is established. The results showed that the prediction model proposed in the study steadily reduced the average absolute error as the number of iterations increase, reaching an average of 0.704, with the loss value stabilising at around 0.6, which is lower than that of the deep neural network prediction model. Compared to the deep neural network prediction model, the normalised discounted cumulative gain is typically 0.01 higher and in terms of hit rate, 0.03 higher. The prediction time of the similarity algorithm is faster than that of the neural network. The mean squared error ranged from a high of 1.29 to a low of 1.19, both lower than other algorithms, and the mean absolute error ranged from a high of 0.56 to a low of 0.54, lower than all other algorithms except the support vector machine algorithm. The average absolute error of the deep neural network resource representation algorithm ranged from a high of 1.46 to a low of 1.45, lower than all other algorithms except the support vector machine algorithm, and the average squared error ranged from a high of 3.43 to a low of 3.24, better than all other algorithms. In conclusion, the model constructed by the study has a good application effect in recommending educational resources, and has a certain promoting effect on the development of online education.

Author 1: Zongkui Wang

Keywords: Intelligent recommendation; deep neural networks; multilayer perceptron; educational resources

Paper 103: Role of Artificial Intelligence and Business Decision Making

Abstract: Artificial Intelligence (AI) has emerged as a transformative technology with profound implications for various sectors, including business. In recent years, AI has revolutionized decision-making processes by providing organizations with advanced analytical capabilities, enabling them to extract valuable insights from vast amounts of data. The application of AI in businesses may force the sector to rely on quicker, less expensive, and more accurate marketing techniques. By utilizing the AI in marketing strategies, a business owner may increase audience reaction and build a strong online brand that can compete with others. In addition to marketing, it has the capacity to remodel a business with fresh concepts. Additionally, it provides solutions for challenging problems, aiding in the enormous business growth. The study's primary goal is to investigate how artificial intelligence and decision-making are deployed in business and tried to explore how AI is being used to enhance decision-making processes and how it is changing business models. The study reveals that the role of artificial intelligence in business decision making is transformative, offering significant advantages in terms of efficiency, accuracy, and innovation. AI-powered systems enable businesses to process and analyze vast amounts of data efficiently, leading to quicker and more informed decision making. Overall, the integration of AI in business decision making has the potential to drive organizational success and shape the future of business practices.

Author 1: Anupama Prasanth

Author 2: Densy John Vadakkan

Author 3: Priyanka Surendran

Author 4: Bindhya Thomas

Keywords: Artificial intelligence; business decision making; efficiency; accuracy; innovation; marketing strategy; machine learning

Paper 104: Unusual Human Behavior Detection System in Real-Time Video Systems

Abstract: Abnormal behavior detection, in terms of importance, has become a necessity in real-time visual systems. The main problem is the ambiguity in the difference between the characteristics of abnormal and normal behavior, which its definition is usually different according to the previous context of images. In this research, three approaches are used. In the first approach, a standard Convolutional Automatic Encoder (CAE) is used. After evaluation, it was found that the standard CAE problem is that it does not take into account the temporal aspect of the image frames sequence. The second method involves automatic encoding to learn the dataset's spatio-temporal structures. In the third approach, the complex LSTM cells are used for further improvement. The outcomes of the test display that the proposed methods have better performance compared to many of the previous conventional methods, and their efficiency in identifying abnormal behavior is very competitive compared to previous methods.

Author 1: Yanbin Bu

Author 2: Ting Chen

Author 3: Hongxiu Duan

Author 4: Mei Liu

Author 5: Yandan Xue

Keywords: Anomaly detection; video sequence; standard Convolutional Automatic Encoder (CAE); spatio-temporal structures; LSTM

Paper 105: A Comprehensive Study of DCNN Algorithms-based Transfer Learning for Human Eye Cataract Detection

Abstract: This study presents a comparative analysis of different deep convolutional neural network (DCNN) architectures, including VGG19, NASNet, ResNet50, and MobileNetV2, with and without data augmentation, for the automatic detection of cataracts in fundus images. Utilizing hybrid architecture models, namely ResNet50-NASNet and ResNet50+MobileNetV2, which combine two state-of-the-art DCNNs, this research demonstrates their superior performance. Specifically, MobileNetV2 and the combined ResNet50+MobileNetV2 outperform other models, achieving an impressive accuracy of 99.00%. By emphasizing the efficacy of diverse datasets and pre-processing techniques, as well as the potential of pretrained DCNN models, this study contributes to accurate cataract diagnosis. Furthermore, the proposed system has the potential to reduce reliance on ophthalmologists, decrease the cost of eye check-ups, and improve accessibility to eye care for a wider population. These findings showcase the successful application of deep learning and image processing techniques in the early detection and treatment of various medical conditions, including cataracts, addressing the needs of individuals with diminished vision through ocular images and innovative hybrid architectures.

Author 1: Omar Jilani Jidan

Author 2: Susmoy Paul

Author 3: Anirban Roy

Author 4: Sharun Akter Khushbu

Author 5: Mirajul Islam

Author 6: S.M. Saiful Islam Badhon

Keywords: Cataract detection; eye disease; ocular images; deep convolutional neural network (DCNN); hybrid architecture

Paper 106: System Dynamics Approach in Supporting The Achievement of The Sustainable Development on MSMEs: A Collection of Case Studies

Abstract: Sustainable development in MSMEs is very important to encourage economic growth, improve the welfare of people, and ensure environmental sustainability. However, achieving sustainability in the MSME sector faces many challenges due to the complex interdependencies and dynamic interactions among various factors. The system dynamics approach makes it possible to model and simulate dynamic feedback loops, time delays, and nonlinear relationships between these factors. This paper provides an overview of the system dynamics approach and its suitability to address the complexities inherent in the MSME sector in its application to sustainable development. It explores the issues faced by MSMEs in achieving sustainable development and how the system dynamics approach models and analyzes the behavior of these MSMEs. These issues cover the dimensions of product development, technology and ICT inclusion, supply chain, business development, financial resources, and organizational support. This study was conducted on several case studies from various industries, namely the steel industry, agro-industry, craft industry, tourism industry, plastic molding, manufacturing, cosmetics, and digital companies; who come from various countries. From this study it was concluded that the system dynamics approach has significant potential to support the achievement of sustainable development in MSMEs, because it allows MSMEs to be able to effectively model and simulate the behavior of various factors that affect their operations, such as resource allocation, environmental impacts, and social considerations; proactively addressing sustainability challenges, adapting to changing market conditions, and contributing to broader socio-economic and environmental objectives.

Author 1: Julia Kurniasih

Author 2: Zuraida Abal Abas

Author 3: Siti Azirah Asmai

Author 4: Agung Budhi Wibowo

Keywords: System dynamics; sustainable development; Micro Small and Medium Enterprises (MSMEs)

Paper 107: A Precise Survey on Multi-agent in Medical Domains

Abstract: Agent technology has provided many opportunities to improve the human standard of life in recent decades, starting from social life and moving on to business intelligence and tackling complicated communication, integration, and analysis challenges. These agents play an important role in human health from diagnosis to treatment. Every day, sophisticated agents and expert systems are being developed for human beings. These agents have made it easier to deal with common diseases and provide high accuracy with less processing time. However, they also have some challenges in their domain, especially when dealing with complex issues. To handle these challenges, the domain has become characterized by distinctive and creative methodologies and architectures. This survey provides a review of medical multi-agent systems, including the typical intelligent agents, their main characteristics and applications, multi-agent systems, and challenges. A classification of multi-agent system applications and challenges is presented, along with references for additional studies. For researchers and practitioners in the field, we intend this paper to be an informative and complete resource on the medical multi-agent system.

Author 1: Arwa Alshehri

Author 2: Fatimah Alshahrani

Author 3: Habib Shah

Keywords: Artificial intelligence; agent systems; multi-agent systems

Paper 108: Technology Adoption and Usage Behaviors in Field Incident Management System Utilization

Abstract: This study utilized the Unified Theory of Acceptance and Use of Technology (UTAUT) model to analyze the adoption and utilization of field incident management system (IMS) in a manufacturing organization. The study specifically focused on the role of behavior as a key factor in the adoption and utilization of incident management system. Data was collected through a survey of employees who had utilized the IMS system and the UTAUT model was used to analyze the data. The results indicated that behavior in the system significantly influenced the adoption and utilization of IMS. The study also found that the UTAUT model provided a useful framework for understanding the adoption and utilization of IMS, particularly the importance of performance expectancy, effort expectancy, social influence, and facilitating conditions. The study provides valuable insights for organizations looking to implement IMS and improve their incident management processes. It highlights the importance of building behavior in the system through appropriate user experience and user training. The findings of this study have important implications for manufacturing organizations seeking to enhance their incident management procedures through the adoption and utilization of IMS.

Author 1: Cory Antonio Buyan

Author 2: Noelyn M. De Jesus

Author 3: Eltimar T. Castro Jr

Keywords: UTAUT; field incident management system (FIMS); regression analysis; user intention and acceptance; system adoption; usage behavior; manufacturing; IMS; effort expectancy; performance expectancy; social influence; facilitating conditions; behavioral intention; ANOVA

Paper 109: Damage Security Intelligent Identification of Wharf Concrete Structures under Deep Learning and Digital Image Technology

Abstract: Artificial Intelligence (AI) technology has quickly developed under the mighty computing power of computers. At this stage, there are many mature non-destructive testing methods in civil engineering, but they are generally only suitable for simple structures and evident damage characteristics. Therefore, it’s necessary for us to investigate the damage identification of wharf concrete structures under deep learning and digital image technology. The article propose a damage detection and localization method based on Neural Network (NN) technology in deep learning and Digital Image Correlation (DIC) to identify internal damage to concrete used for wharf construction. Firstly, the identification model of concrete structure is constructed using NN technology. Then, structural damage identification of concrete is further investigated using DIC. Finally, relevant experiments are designed to verify the effect of the model. The results show that: (1) The damage model of concrete structure constructed by NN technology has high convergence and stability and can control the test error well. (2) The image output by the DIC equipment is processed and input into the NN. The errors of the various parameters of different concretes can be within the acceptable range. This paper aims to provide good ideas and references for follow-up structural health monitoring and other topics and has significant engineering application value.

Author 1: Jinbo Zhu

Author 2: Yuesong Li

Author 3: Pengrui Zhu

Keywords: Structural damage identification; deep learning; neural network; digital image; concrete

Paper 110: Application of Top-N Rule-based Optimal Recommendation System for Language Education Content based on Parallel Computing

Abstract: In recent years personalized recommendation services have been applied to many areas of society, typically in the fields of e-commerce, short videos and so on. In response to the serious performance problems of the current online language education platform content recommendation, so in the face of the above opportunities and challenges, this paper designs a new online English education model to allow university students to get a full and more three-dimensional training of English language learning. Based on the MU platform, this paper obtains data from the platform and uses crawler technology to sample and standardize the learning resources for online education. Then user information, such as explicit and implicit ratings of courses, is selected as the main basis for training a user interest preference model. Immediately afterwards, a PRF algorithm combining data parallelism and task parallelism optimization was executed and implemented on Apeche Spark to provide some optimization of data accuracy and content recommendation methods. Finally, the top-N recommendation rule is used to propose a dynamic evolutionary process of identifying students’ preferences or learning habits through the results of previous data analysis, so as to make more accurate course content recommendations and learning content guidance for students’ English learning. The online three-dimensional teaching model proposed in this paper focuses more on time-series research than traditional algorithms, and can more accurately capture the dynamic changes in students’ learning abilities.

Author 1: Nan Hu

Keywords: Data parallel computing; cloud computing; data crawlers; top-N rules; PRF algorithm

Paper 111: A Novel ML Approach for Computing Missing Sift, Provean, and Mutassessor Scores in Tp53 Mutation Pathogenicity Prediction

Abstract: Cancer is often caused by missense mutations, where a single nucleotide substitution leads to an amino acid change and affects protein function. This study proposes a novel machine learning (ML) approach to calculate missing values in the tp53 database for three computational methods: SIFT, Provean, and Mutassessor scores. The computed values are compared with those obtained from the imputation method. Using these values, an ML classification model trained on 80,406 samples achieves an accuracy of 85%, while the impute method achieves 75%. The scores and statistics are used to classify samples into five classes: Benign, likely pathogenic, possibly pathogenic, pathogenic, and a variant of uncertain significance. Additionally, a comparative analysis is conducted on 58,444 samples, evaluating six ML techniques. The accuracy obtained by each of these mentioned in mentioned alongside the algorithm: logistic regression (89%), k-nearest neighbor (99%), decision tree (95%), random forest (99.8%), support vector machine with the polynomial kernel (91%), support vector machine with RBF kernel (84%), and deep neural networks (98.2%). These results demonstrate the effectiveness of the proposed ML approach for pathogenicity prediction.

Author 1: Rashmi Siddalingappa

Author 2: Sekar Kanagaraj

Keywords: Decision tree (DT); deep neural networks (DNN); imputation; k-nearest neighbor (KNN); logistic regression (LR); missense mutations; Mutassessor; pathogenicity; Provean; random forest (RF); SIFT; support vector machine (SVM)

Paper 112: An Empirical Deep Learning Approach for Arabic News Classification

Abstract: In this paper, we tackle the problem of Arabic news classification. A dataset of 5,000 news articles from various Saudi Arabian news sources were gathered, classified into six categories: business, entertainment, health, politics, sports, and technology. We conducted experiments using different pre-processing techniques, word embeddings, and deep learning architectures, including convolutional neural networks (CNNs) and long short-term memory (LSTM) networks, as well as a hybrid CNN-LSTM model. Our proposed model achieved an accuracy of 93.15, outperforming other models in terms of accuracy. Moreover, our model is evaluated on other Arabic news datasets and obtained competitive results. Our approach demonstrates the effectiveness of deep learning methods in Arabic news classification and emphasizes the significance of careful selection of preprocessing techniques, word embeddings, and deep learning architectures.

Author 1: Roobaea Alroobaea

Keywords: Deep learning (DL); machine-learning (ML); convolutional neural networks (CNNs); long short-term memory (LSTM)

Paper 113: Application Methods of Image Design Based on Virtual Reality and Interaction

Abstract: The continuous improvement of virtual reality and interactive technology has led to a broader and deeper application in related fields, especially image design. In image design, creating usage scenarios for portable interactive experience products based on virtual reality and interactive technology can optimize and improve key parameters for real 3D techniques, thereby building a more comprehensive image design. This article constructs a three-dimensional image model of marine organisms and scenarios based on multi-sensory interactive interface generation technology and information fusion optimization ANNs-DS algorithm, targeting the image scenarios of product design. The relevant model information and parameter changes are analyzed. The results indicate that in the process of multi-sensory interface interactive image design, the virtual reality image design implemented using ANNs-DS information fusion algorithm can enhance participants' multi-sensory visual experience of the interactive interface. The reasonable degree between objects in the interactive interface and the scene space image is basically within the range of 0.85-0.95. The fluency in different scenarios can be significantly improved. Therefore, virtual reality and interactive technology have laid the foundation for developing interactive image design.

Author 1: Shasha Mao

Keywords: Virtual reality; interactive; ANNs-DS information fusion algorithm; image design

Paper 114: Video Surveillance Vehicle Detection Method Incorporating Attention Mechanism and YOLOv5

Abstract: With the rising number of vehicle ownership nationwide and the consequent increase in traffic accidents, vehicle detection for traffic surveillance video is an effective method to reduce traffic accidents. However, existing video surveillance vehicle detection methods suffer from high computational load, low accuracy, and excessive reliance on large-scale computing servers. Therefore, the research will try to fuse coordinate attention mechanism to improve YOLOv5 network, choose lightweight YOLOv5s for image recognition, and use K-means algorithm to modify the aiming frame according to the characteristics of vehicle detection; meanwhile, in order to get more accurate results, coordinate attention mechanism algorithm, which is also a lightweight algorithm, is inserted into YOLOv5s for improvement, so that the designed The lightweight vehicle detection model can be run on embedded devices. The measurement experiments show that the YOLOv5+CA model completes convergence when the iterations exceed 100, and the localization loss and confidence loss gradually stabilize at 0.002 and 0.028, and the classification loss gradually stabilizes at 0.017. Comparing YOLOv5+CA with SSD algorithm, ResNet-101 algorithm and RefineDet algorithm, YOLOv5 +CA detection accuracy is better than other algorithms by about 9%, and the accuracy can be approximated to 1.0 at a confidence level of 0.946. The experimental results show that the research design provides higher accuracy and high computational efficiency for video surveillance vehicle detection, and can better provide reference value and reference methods for video surveillance vehicle detection and operation management.

Author 1: Yi Pan

Author 2: Zhu Zhao

Author 3: Yan Hu

Author 4: Qing Wang

Keywords: Attention mechanism; YOLOv5; vehicle detection; image recognition; deep learning

Paper 115: Stroke Risk Prediction: Comparing Different Sampling Algorithms

Abstract: Stroke is a serious disease that has a significant impact on the quality of life and safety of patients. Accurately predicting stroke risk is of great significance for preventing and treating stroke. In the past few years, machine learning methods have shown potential in predicting stroke risk. However, due to the imbalance of stroke data and the challenges of feature selection and model selection, stroke risk prediction still faces some difficulties.This article aims to compare the performance differences between different sampling algorithms and machine learning methods in stroke risk prediction. This study used the over-sampling algorithm (Random Over Sampling and SMOTE), the under-sampling algorithm (Random Under Sampling and ENN), and the hybrid sampling algorithm (SMOTE-ENN), and combined them with common machine learning methods such as K-Nearest Neighbors, Logistic Regression, Decision Tree and Support Vector Machine to build the prediction model.Through the analysis of experimental results, and found that the SMOTE combined with the LR model showed good performance in stroke risk prediction, with a high F1 score. In addition, this study found that the overall performance of the undersampling algorithm is better than that of the oversampling and hybrid sampling algorithms.These research results provide useful references for predicting stroke risk and provide a foundation for further research and application. Future research can continue to explore more sampling algorithms, machine learning methods, and feature engineering techniques to further improve the accuracy and interpretability of stroke risk prediction and promote its application in clinical practice.

Author 1: Qiuyang Yin

Author 2: Xiaoyan Ye

Author 3: Binhua Huang

Author 4: Lei Qin

Author 5: Xiaoying Ye

Author 6: Jian Wang

Keywords: Stroke prediction; data mining; machine learning; unbalanced data; sampling algorithms; classification algorithms

Paper 116: Comparative Analysis using Various Performance Metrics in Imbalanced Data for Multi-class Text Classification

Abstract: Precision, Recall, and F1-score are metrics that are often used to evaluate model performance. Precision and Recall are very important to consider when the data is balanced, but in the case of unbalanced data the F1-score is the most important metric. To find out the importance of these metrics, a comparative analysis is needed in order to determine which metric is appropriate for the data being analyzed. This study aims to perform a comparative analysis of various evaluation metrics on unbalanced data in multi-class text classification. This study uses an unbalanced multi-class text dataset including: association, negative, cause of disease, and treatment of disease. This study involves five classifiers as algorithm-level approach, namely: Multinomial Naive Bayes, K-Nearest Neigbors, Support Vector Machine, Random Forest, and Long Short-Term Memory. Mean-while, data-level approach, this study involves under sampling, over sampling, and synthetic minority oversampling technique. Several evaluation metrics used to evaluate model performance include Precision, Recall, and F1-score. The results show that the most suitable evaluation metric for use on unbalanced data depends on the purpose of use and the desired priority, including the classifier that is suitable for handling multi-class assignments on unbalanced data. The results of this study can assist practitioners in selecting evaluation metrics that are in accordance with the goals and application needs of multi-class text classification.

Author 1: Slamet Riyanto

Author 2: Imas Sukaesih Sitanggang

Author 3: Taufik Djatna

Author 4: Tika Dewi Atikah

Keywords: Imbalanced data; undersampling; oversampling; smote; machine learning

Paper 117: Multi-objective Task Scheduling Optimization Based on Improved Bat Algorithm in Cloud Computing Environment

Abstract: In cloud computing environments, task completion time and virtual machine load balance are two critical issues that need to be addressed. To solve these problems, this paper proposes a Multi-objective Optimization Mutate Discrete Bat Algorithm (MOMDBA) that improves upon the traditional Bat algorithm (BA). The MOMDBA algorithm introduces a mutation factor and mutation inertia weight during the global optimization process to enhance the algorithm’s global search ability and convergence speed. Additionally, the local optimization logic is optimized according to the characteristics of cloud computing task scenarios to improve the degree of load balancing of virtual machines. Simulation experiments were conducted using CloudSim to evaluate the algorithm’s performance, and the results were compared with other scheduling algorithms. The results of our experiments indicate that when the cost difference between algorithms is within 4.47%, MOMDBA can significantly outperform other methods. Specifically, compared to PSO, GA, and LBACO, our algorithm reduces makespan by 56.26%, 59.87%, and 25.26%, respectively, while also increasing the degree of load balancing by 93.87%, 75.92%, and 39.13%, respectively. These findings demonstrate the superior performance of MOMDBA in optimizing task scheduling and load balancing.

Author 1: Dakun Yu

Author 2: Zhongwei Xu

Author 3: Meng Mei

Keywords: Cloud computing; task scheduling; optimization; bat algorithm; meta-heuristics

Paper 118: An Adaptive Testcase Recommendation System to Engage Students in Learning: A Practice Study in Fundamental Programming Courses

Abstract: This paper proposes a testcase recommendation system (TRS) to assist beginner-level learners in introductory programming courses with completing assignments on a learning management system (LMS). These learners often struggle to generate complex testcases and handle numerous code errors, leading to disengaging their attention from the study. The proposed TRS addresses this problem by applying the recommendation system using singular value decomposition (SVD) and the zone of proximal development (ZPD) to provide a small and appropriate set of testcases based on the learner’s ability. We implement this TRS to the university-level Fundamental Programming courses for evaluation. The data analysis has demonstrated that TRS significantly increases student interactions with the system.

Author 1: Tien Vu-Van

Author 2: Huy Tran

Author 3: Thanh-Van Le

Author 4: Hoang-Anh Pham

Author 5: Nguyen Huynh-Tuong

Keywords: Testcases recommendation system (TRS); learning management system (LMS); zone of proximal development (ZPD); singular value decomposition (SVD)

Paper 119: Brain Tumor Semantic Segmentation using Residual U-Net++ Encoder-Decoder Architecture

Abstract: Image segmentation is considered one of the essential tasks for extracting useful information from an image. Given the brain tumor and its consumption of medical resources, the development of a deep learning method for MRI to segment the brain tumor of patients’ MRI is illustrated here. Brain tumor segmentation technique is crucial in detecting and treating MRI brain tumors. Furthermore, it assists physicians in locating and measuring tumors and developing treatment and rehabilitation programs. The residual U-Net++ encoder-decoder-based architecture is designed as the primary network, and it is an architecture that is hybridized between ResU-Net and U-Net++. The proposed Residual U-Net++ is applied to MRI brain images for the most recent and well-known global benchmark challenges: BraTS 2017, BraTS 2019, and BraTS 2021. The proposed approach is evaluated based on brain tumor MRI images. The results with the BraST 2021 dataset with a dice similarity coefficient (DSC) is 90.3%, sensitivity is 96%, specificity is 99%, and 95% Hausdorff distance (HD) is 9.9. With the BraST 2019 dataset, a DSC is 89.2%, sensitivity is 96%, specificity is 99%, and HD is 10.2. With the BraST 2017 dataset, a DSC is 87.6%, sensitivity is 94%, specificity is 99%, and HD is 11.2. Furthermore, Residual U-Net++ outperforms the standard brain tumor segmentation approaches. The experimental results indicated that the proposed method is promising and can provide better segmentation than the standard U-Net. The segmentation improvement could help radiologists increase their radiologist segmentation accuracy and save time by 3%.

Author 1: Mai Mokhtar

Author 2: Hala Abdel-Galil

Author 3: Ghada Khoriba

Keywords: Brain tumor segmentation; medical image segmentation; BraTS; U-Net; U-Net++; residual network

Paper 120: Multi-dimensional Data Aggregation Scheme Supporting Fault-Tolerant Mechanism in Smart Grid

Abstract: With the large-scale deployment of smart grids, the scheme of smart grid data aggregation has gradually enriched in recent years. Based on the principle of protecting user privacy, existing schemes usually choose to introduce a trusted third party (TTP) to participate in the collaboration. However, this also increases the risk of privacy exposure as the attacker can target the TTP which provides services to smart grid operators. In addition, many existing schemes do not take into account the operational requirements of smart meters in case of failure. Furthermore, some schemes ignore the control center’s demand for analyzing multi-dimensional data, which causes a lot of inconvenience in actual operation. Therefore, a fault-tolerant multi-dimensional data aggregation scheme is proposed in this paper. We have constructed a scheme without TTP participation in collaboration, and also meet the following two requirements. The scheme not only ensures the normal operation of the system when the smart meter fails but also meets the requirements of the control center for multi-dimensional data analysis. Security analysis shows that the proposed scheme can resist external attack, internal attack, and collusion attack. The experimental results show that the proposed scheme improves the fault tolerance and reduces the computational cost compared with the existing schemes.

Author 1: Yong Chen

Author 2: Feng Wang

Author 3: Li Xu

Author 4: Zhongming Huang

Keywords: Cryptography; fault tolerance; privacy; multi-dimensional data aggregation; encryption; smart grid

Paper 121: SbChain+: An Enhanced Snowball-Chain Approach for Detecting Communities in Social Graphs

Abstract: In this paper, we present snowball-chain (SbChain+) approach, which is an improved version of SbChain community detection method in terms of precision with which communities are identified in a social graph. It exploits the topology of a social graph in terms of the connections of a node, i.e., its degree centrality, betweenness centrality and the number of links within its neighborhood defined by the local clustering coefficient. Two different functions have been used to identify neighbors for a given node. Hence, two approaches have been discussed with their pros and cons. In general, SbChain+ takes a social graph as an input and aims to identify communities around the core nodes in the underlying network. The core nodes are expected to have a high degree and have densely connected neighbors and guides in identifying cliques from the graph. The proposed approach takes its inspiration from snowball sampling technique and keeps merging the nodes with their neighboring nodes based on certain criteria to form snowballs. One of the functions discussed (SbChain+(i)) uses a hyperparameter, λ for merging snowballs which further leads to the formation of communities. This hyperparameter also helps in achieving the desired level of coarseness in the communities, and it can be adjusted to fine tune the identified communities. While the second function (SbChain+(ii)) uses an average out degree function to merge snowballs. The modularity values are calculated at each level of the dendrogram formed by combining nodes and snowballs to decide an appropriate cut for community determination. SbChain+ is empirically evaluated using these two different functions over both real-world and LFR-benchmark datasets and results are evaluated on modularity and normalized mutual information. The aim of this study is to improve upon the previously discussed technique (SbChain) and to study the use of hyperparameter, i.e., the performance of a technique with or without the hyperparameter.

Author 1: Jayati Gulati

Author 2: Muhammad Abulaish

Keywords: Clique; clustering; community detection; graph mining; snowball sampling; social network analysis

Paper 122: PaddyNet: An Improved Deep Convolutional Neural Network for Automated Disease Identification on Visual Paddy Leaf Images

Abstract: Timely disease diagnosis in paddy is fundamental to preventing yield losses and ensuring an adequate supply of rice for a rapidly rising worldwide population. Recent advancements in deep learning have helped overcome the limitations of unsuper-vised learning methods. This paper proposes a novel PaddyNet model for enhanced accuracy in paddy leaf disease detection. The PaddyNet model, developed using 17 layers, captures and models patterns of different disease symptoms present in paddy leaf images. The effectiveness of the novel model is verified by applying a large dataset comprising 16,225 paddy leaf datasets across 13 classes, including a normal class and 12 disease classes. The performance results show that the new PaddyNet model classifies paddy leaf disease images effectively with 98.99%accuracy and a dropout value of 0.4.

Author 1: Petchiammal A

Author 2: Murugan D

Author 3: Briskline Kiruba S

Keywords: Image annotation; data augmentation; deep learn-ing; paddy leaf disease detection; paddyNet

Paper 123: Unmanned Aerial Vehicle-based Applications in Smart Farming: A Systematic Review

Abstract: On one hand, the emergence of cutting-edge technologies like AI, Cloud Computing, and IoT holds immense potential in Smart Farming and Precision Agriculture. These technologies enable real-time data collection, including high-resolution crop imagery, using Unmanned Aerial Vehicles (UAVs). Leveraging these advancements can revolutionize agriculture by facilitating faster decision-making, cost reduction, and increased yields. Such progress aligns with precision agriculture principles, optimizing practices for the right locations, times, and quantities. On the other hand, integrating UAVs in Smart Farming faces obstacles related to technology selection and deployment, particularly in data acquisition and image processing. The relative novelty of UAV utilization in Precision Agriculture contributes to the lack of standardized workflows. Consequently, the widespread adoption and implementation of UAV technologies in farming practices are hindered. This paper addresses these challenges by conducting a comprehensive review of recent UAV applications in Precision Agriculture. It explores common applications, UAV types, data acquisition techniques, and image processing methods to provide a clear understanding of each technology’s advantages and limitations. By gaining insights into the advantages and challenges associated with UAV-based applications in Precision Agriculture, this study aims to contribute to the development of standardized workflows and improve the adoption of UAV technologies.

Author 1: El Mehdi Raouhi

Author 2: Mohamed Lachgar

Author 3: Hamid Hrimech

Author 4: Ali Kartit

Keywords: Artificial intelligence; internet of things; sensor; big data; cloud; unmanned aerial vehicle; smart farming

Paper 124: Advanced Night time Object Detection in Driver-Assistance Systems using Thermal Vision and YOLOv5

Abstract: Driver-assistance systems have become an indispensable component of modern vehicles, serving as a crucial element in enhancing safety for both drivers and passengers. Among the fundamental aspects of these systems, object detection stands out, posing significant challenges in low-light scenarios, particularly during nighttime. In this research paper, we propose an innovative and advanced approach for detecting objects during nighttime in driver-assistance systems. Our proposed method leverages thermal vision and incorporates You Only Look Once version 5 (YOLOv5), which demonstrates promising results. The primary objective of this study is to comprehensively evaluate the performance of our model, which utilizes a combination of stochastic gradient descent (SGD) and Adam optimizer. Moreover, we explore the impact of different activation functions, including SiLU, ReLU, Tanh, LeakyReLU, and Hardswish, on the efficiency of nighttime object detection within a driver assistance system that utilizes thermal imaging. To assess the effectiveness of our model, we employ standard evaluation metrics including precision, recall, and mean average precision (mAP), commonly used in object detection systems.

Author 1: Hoang-Tu Vo

Author 2: Luyl-Da Quach

Keywords: Driver-assistance systems; object detection; nighttime object detection; thermal vision; YOLOv5

Paper 125: Hate Speech Detection in Bahasa Indonesia: Challenges and Opportunities

Abstract: This study aims to provide an overview of the current research on detecting abusive language in Indonesian social media. The study examines existing datasets, methods, and challenges and opportunities in this field. The research found that most existing datasets for detecting abusive language were collected from social media platforms such as Twitter, Facebook, and Instagram, with Twitter being the most commonly used source. The study also found that hate speech is the most researched type of abusive language. Various models, including traditional machine learning and deep learning approaches, have been implemented for this task, with deep learning models showing more competitive results. However, the use of transformer-based models is less popular in Indonesian hate speech studies. The study also emphasizes the importance of exploring more diverse phenomena, such as islamophobia and political hate speech. Additionally, the study suggests crowdsourcing as a potential solution for the annotation approach for labeling datasets. Furthermore, it encourages researchers to consider code-mixing issues in abusive language datasets in Indonesia, as it could improve the overall model performance for detecting abusive language in Indonesian data. The study also suggests that the lack of effective regulations and the anonymity afforded to users on most social networking sites, as well as the increasing number of Twitter users in Indonesia, have contributed to the rising prevalence of hate speech in Indonesian social media. The study also notes the importance of considering code-mixed language, out-of-vocabulary words, grammatical errors, and limited context when working with social media data.

Author 1: Endang Wahyu Pamungkas

Author 2: Divi Galih Prasetyo Putri

Author 3: Azizah Fatmawati

Keywords: Abusive language; hate speech detection; machine learning; social media

Paper 126: MC-ABAC: An ABAC-based Model for Collaboration in Multi-Cloud Environment

Abstract: Collaborative systems allow a group of organizations to collaborate and complete shared tasks through distributed platforms. Organizations who collaborate often leverage cloud-based solutions to outsource their data and to benefit from the cloud capabilities. During such collaborations, tenants require access to and utilize resources held by other collaborating tenants, which are hosted across multiple cloud providers. Ensuring access control in a cloud-based collaborative application is a crucial problem that needs to be addressed, particularly in a multi-cloud environment. This paper presents the Multi-Cloud ABAC: MC-ABAC model, an extension of the ABAC: Attribute Based Access Control model, suitable for ensuring secure collaboration and cross-tenant access in a multi-cloud environment. The MC-ABAC introduces the notions of tenant, cloud customer and cloud service provider as fundamental entities within the model. Additionally, it incorporates multiple trust relations to enable collaboration and resource sharing among tenants in the multi-cloud environment. To demonstrate its feasibility, we have implemented the MC-ABAC model using Python technology.

Author 1: Mohamed Amine Madani

Author 2: Abdelmounaim Kerkri

Author 3: Mohammed Aissaoui

Keywords: ABAC model; multi-tenant; multi-cloud; collaboration; trust

Paper 127: Type 2 Diabetes Mellitus: Early Detection using Machine Learning Classification

Abstract: Type 2 Diabetes Mellitus (T2DM) is a growing global health problem that significantly impacts patient’s quality of life and longevity. Early detection of T2DM is crucial in preventing or delaying the onset of its associated complications. This study aims to evaluate the use of machine learning algorithms for the early detection of T2DM. A classification model is developed using a dataset of patients diagnosed with T2DM and healthy controls, incorporating feature selection techniques. The model will be trained and tested on machine learning algorithms such as Logistic Regression, K-Nearest Neighbors, Decision Trees, Random Forest, and Support Vector Machines. The results showed that the Random Forest algorithm achieved the highest accuracy in detecting T2DM, with an accuracy of 98%. This high accuracy rate highlights the potential of machine learning algorithms in early T2DM detection and the importance of incorporating such methods in the clinical decision-making process. The findings of this study will contribute to the development of a more efficient precision medicine screening process for T2DM that can help healthcare providers detect the disease at its earliest stages, leading to improved patient outcomes.

Author 1: Gowthami S

Author 2: Venkata Siva Reddy

Author 3: Mohammed Riyaz Ahmed

Keywords: Diabetes Mellitus Type II; feature selection; machine learning methods; precision medicine

Paper 128: A Novel Classification Approach for Grape Leaf Disease Detection Based on Different Attention Deep Learning Techniques

Abstract: Preventing and controlling grape diseases is essential for a good grape harvest. With the help of “single shot multi-box detectors”, “faster region based convolutional neural networks”, & “You only look once-X,” the study improved grape leaf disease detection accuracy with effective attention mechanisms, which includes convolutional block attention module, squeeze & excitation networks, & efficient channel attention. The various attention techniques helped to emphasize important features while reducing the impact of irrelevant ones, which ultimately improved the precision of the models and allowed for real-time performance. As a result of examining the optimal models from the three types, it was found that the Faster (R-CNN) model had a lower precision value, while You only look once-X and SSD with various attention techniques required the fewest parameters with the highest precision, with the best real-time performance. In addition to providing insights into grape diseases & symptoms in automated agricultural production, this study provided valuable insights into grape leaf disease detection.

Author 1: S Phani Praveen

Author 2: Rajeswari Nakka

Author 3: Anuradha Chokka

Author 4: Venkata Nagaraju Thatha

Author 5: Sai Srinivas Vellela

Author 6: Uddagiri Sirisha

Keywords: Grape leaves; faster region-based convolutional neural networks; you only look once (x); single shot detection attention techniques

Paper 129: Predicting At-Risk Students’ Performance Based on LMS Activity using Deep Learning

Abstract: It is of great importance for Higher Education (HE) institutions to continuously work on detecting at-risk students based on their performance during their academic journey with the purpose of supporting their success and academic advancement. This is where Learning Analytics (LA) representing learners’ behaviour inside the Learning Management Systems (LMS), Educational Data Mining (EDM), and Deep Learning (DL) techniques come into play as an academic sustainable pipeline, which can be used to extract meaningful predictions of the learners’ future performance based on their online activity. Thus, the aim of this study is to implement a supervised learning approach which utilizes three artifcial neural networks (vRNN, LSTM, and GRU) to develop models that can classify students’ final grade as Pass or Fail based on a number of LMS activity indicators; more precisely, detect failed students who are actually the ones susceptible to risk. The three models alongside a baseline MLP classifier have been trained on two datasets (ELIA 101- 1, and ELIA 101-2) illustrating the LMS activity and final assessment grade of 3529 students who enrolled in an English Foundation-Year course (ELIA 101) taught at King Abdulaziz University (KAU) during the first and second semesters of 2021. Results indicate that though all of the three DL models performed better than the MLP baseline, the GRU model achieved the highest classification accuracy on both datasets: 93.65% and 98.90%, respectively. As regards predicting at-risk students, all of the three DL models achieved an = 81% Recall values, with notable variation of performance depending on the dataset, the highest being the GRU on the ELIA 101-2.

Author 1: Amnah Al-Sulami

Author 2: Miada Al-Masre

Author 3: Norah Al-Malki

Keywords: Predict at-risk student; artificial neural network; learning management system; and educational data mining

Paper 130: Parameter Identification of a Multilayer Perceptron Neural Network using an Optimized Salp Swarm Algorithm

Abstract: Effort estimation in software development (SEE) is a crucial concern within the software engineering domain, as it directly impacts cost estimation, scheduling, staffing, planning, and resource allocation accuracy. In this scientific article, the authors aim to tackle this issue by integrating machine learning (ML) techniques with metaheuristic algorithms in order to raise prediction accuracy. For this purpose, they employ a multilayer perceptron neural network (MLP) to perform the estimation for SEE. Unfortunately, the MLP network has numerous drawbacks as well, including weight dependency, rapid convergence, and accuracy limits. To address these issues, the SSA Algorithm is employed to optimize the MLP weights and biases. Simultaneously, the SSA algorithm has shortcomings in some aspects of the search mechanisms as well, such as rapid convergence and being susceptible to the local optimal trap. As a result, the genetic algorithm (GA) is utilized to address these shortcomings through fine-tuning its parameters. The main objective is to develop a robust and reliable prediction model that can handle a wide range of SEE problems. The developed techniques are tested on twelve benchmark SEE datasets to evaluate their performance. Furthermore, a comparative analysis with state-of-the-art methods is conducted to further validate the effectiveness of the developed techniques. The findings demonstrate that the developed techniques surpass all other methods in all benchmark problems, affirming their superiority.

Author 1: Mohamad Al-Laham

Author 2: Salwani Abdullah

Author 3: Mohammad Atwah Al-Ma’aitah

Author 4: Mohammed Azmi Al-Betar

Author 5: Sofian Kassaymeh

Author 6: Ahmad Azzazi

Keywords: Software development effort estimation; machine learning; multilayer perceptron neural network; salp swarm algorithm; genetic algorithm

Paper 131: A Novel Method for Diagnosing Alzheimer’s Disease from MRI Scans using the ResNet50 Feature Extractor and the SVM Classifier

Abstract: Alzheimer’s disease (AD), a chronic neurodegenerative brain disorder, caused by the accumulation of abnormal proteins called amyloid, is one of the prominent causes of mortality worldwide. Since there is a scarcity of experienced neurologists, manual diagnosis of AD is very time-consuming and error-prone. Hence, automatic diagnosis of AD draws significant attention nowadays. Machine learning (ML) algorithms such as deep learning are widely used to support early diagnosis of AD from magnetic resonance imaging (MRI). However, they provide better accuracy in binary classification, which is not the case with multi-class classification. On the other hand, AD consists of a number of early stages, and accurate detection of them is necessary. Hence, this research focuses on how to support the multi-stage classification of AD particularly in its early stage. After the MRI scans have been preprocessed (through median filtering and watershed segmentation), benchmark pre-trained convolutional neural network (CNN) models (AlexNet, VGG16, VGG19, ResNet18, ResNet50) carry out automatic feature extraction. Then, principal component analysis is used to optimize features. Conventional machine learning classifiers (Decision Tree, K-Nearest Neighbors, Support Vector Machine, Linear Programming Boost, and Total Boost) are deployed using the optimized features for staging AD. We have exploited the Alzheimer’s disease Neuroimaging Initiative(ADNI) data set consisting of AD, MCIs (MCI), and cognitive normal (CN) classes of images. In our experiment, the SVM classifier performed better with the extracted ResNet50 features, achieving multi-class classification accuracy of 99.78% during training, 99.52% during validation, and 98.71% during testing. Our approach is distinctive because it combines the advantages of deep feature extractors, conventional classifiers, and feature optimization.

Author 1: Farhana Islam

Author 2: Md. Habibur Rahman

Author 3: Nurjahan

Author 4: Md. Selim Hossain

Author 5: Samsuddin Ahmed

Keywords: Alzheimer’s disease; brain images; machine learning; deep learning; brain disorder; ADNI dataset

Paper 132: Uncertainty-Aware Traffic Prediction using Attention-based Deep Hybrid Network with Bayesian Inference

Abstract: Traffic congestion has an adverse impact on the economy and quality of life and thus accurate traffic flow forecasting is critical for reducing congestion and enhancing transportation management. Recently, hybrid deep-learning approaches show promising contributions in prediction by handling various dynamic traffic features. Existing methods, however, frequently neglect the uncertainty associated with traffic estimates, resulting in inefficient decision-making and planning. To overcome these issues, this research presents an attention-based deep hybrid network with Bayesian inference. The suggested approach assesses the uncertainty associated with traffic projections and gives probabilistic estimates by applying Bayesian inference. The attention mechanism improves the ability of the model to detect unexpected situations that disrupt traffic flow. The proposed method is tested using real-world traffic data from Dhaka city, and the findings show that it outperforms than other cutting-edge approaches when used with real-world traffic statistics.

Author 1: Md. Moshiur Rahman

Author 2: Abu Rafe Md Jamil

Author 3: Naushin Nower

Keywords: Traffic flow prediction; uncertainty; deep learning; Bayesian inference; Dhaka city

Paper 133: An Enhanced Variational AutoEncoder Approach for the Purpose of Deblurring Bangla License Plate Images

Abstract: Automated License Plate Detection and Recognition (ALPDR) is a well-studied area of computer vision and a crucial activity in a variety of applications, including surveillance, law enforcement, and traffic management. Such a system plays a crucial role in the investigation of vehicle-related offensive activities. When an input image or video frame travels through an ALPDR system for license plate detection, the detected license plate is frequently blurry due to the fast motion of the vehicle or low-resolution input. Images of license plates that are blurred or distorted can reduce the accuracy of ALPDR systems. In this paper, a novel Variational AutoEncoder(VAE) architecture is proposed for deblurring license plates. In addition, a dataset of obscured license plate images and corresponding ground truth images is proposed and used to train the novel VAE model. This dataset comprises 3788 image pairs, in which the train, test, and validation set contains 2841, 568, and 379 pairs of images respectively. Upon completion of the training process, the model undergoes an evaluation procedure utilizing the validation set, where it achieved an SSIM value of 0.934 and a PSNR value of 32.41. In order to assess the efficacy of our proposed VAE model, a comparison with contemporary deblurring techniques is pre-sented in the results section. In terms of both quantitative metrics and the visual quality of the deblurred images, the experimental results indicate that our proposed method outperforms the other state-of-the-art deblurring methods. Therefore, it enhances the precision and dependability of an ALPDR system.

Author 1: Md. Siddiqure Rahman Tusher

Author 2: Nakiba Nuren Rahman

Author 3: Shabnaz Chowdhury

Author 4: Anika Tabassum

Author 5: Md. Akhtaruzzaman Adnan

Author 6: Rashik Rahman

Author 7: Shah Murtaza Rashid Al Masud

Keywords: Image deblur; bangla license plate deblur; Variational AutoEncoder (VAE); computer vision

Paper 134: Facial Image Generation from Bangla Textual Description using DCGAN and Bangla FastText

Abstract: The synthesis of facial images from textual descriptions is a relatively difficult subfield of text-to-image synthesis. It is applicable in various domains like Forensic Science, Game Development, Animation, Digital Marketing, and Metaverse. However, no work was found that generates facial images from textual descriptions in Bangla; the 5th most spoken language in the world. This research introduces the first-ever system to generate facial images from Bangla textual descriptions. The proposed model comprises two fundamental constituents, namely a textual encoder, and a Generative Adversarial Network(GAN). The text encoder is a pre-trained Bangla text encoder named Bangla FastText which is employed to transform Bangla text into a latent vector representation. The utilization of Deep Convolutional GAN (DCGAN) allows for the generation of face images that correspond to text embedding. Furthermore, a Bangla version of the CelebA dataset, CelebA Bangla is created for this study to develop the proposed system. CelebA Bangla contains images of celebrities, their corresponding annotated Bangla facial attributes and Bangla Textual Descriptions generated using a novel description generation algorithm. The proposed system attained a Fr´echet Inception Distance (FID) score of 126.708, Inception Score(IS) of 12.361, and Face Semantic Distance(FSD) of 20.23. The novel text embedding strategy used in this study outperforms prior work. A thorough qualitative and quantitative analysis demonstrates the superior performance of the proposed system over other experimental systems.

Author 1: Noor Mairukh Khan Arnob

Author 2: Nakiba Nuren Rahman

Author 3: Saiyara Mahmud

Author 4: Md. Nahiyan Uddin

Author 5: Rashik Rahman

Author 6: Aloke Kumar Saha

Keywords: Bangla Text-to-Face Synthesis; Natural Language Processing(NLP); Computer Vision(CV); GAN; Text encoders

Paper 135: Microbial Biomarkers Identification for Human Gut Disease Prediction using Microbial Interaction Network Embedded Deep Learning

Abstract: Human gut microorganisms are crucial in regulating the immune system. Disruption of the healthy relationship between the gut microbiota and gut epithelial cells leads to the development of diseases. Inflammatory Bowel Disease (IBD) and Colorectal Cancer (CRC) are gut-related disorders with complex pathophysiological mechanisms. With the massive availability of microbiome data, computer-aided microbial biomarker discovery for IBD and CRC is becoming common. However, microbial interactions were not considered by many of the existing biomarker identification methods. Hence, in this study, we aim to construct a microbial interaction network (MIN). The MIN accounts for the associations formed and interactions among microbes and hosts. This work explores graph embedding feature selection through the construction of a sparse MIN using MAGMA embedded into a deep feedforward neural network (DFNN). This aims to reduce dimensionality and select prominent features that form the disease biomarkers. The selected features are passed through a deep forest classifier for disease prediction. The proposed methodology is experimentally cross-validated (5-fold) with different classifiers, existing works, and different models of MIN embedded in DFNN for the IBD and CRC datasets. Also, the selected biomarkers are verified against biological studies for the IBD and CRC datasets. The highest achieved AUC, accuracy, and f1-score are 0.863, 0.839, and 0.897, respectively, for the IBD dataset and 0.837, 0.768, and 0.757, respectively, for the CRC dataset. As observed, the proposed method is successful in selecting a subset of informative and prominent biomarkers for IBD and CRC.

Author 1: Anushka Sivakumar

Author 2: Syama K

Author 3: J. Angel Arul Jothi

Keywords: Biomarker discovery; microbial interaction network; graph embedding feature selection; inflammatory bowel disease; colorectal cancer

Paper 136: Software Vulnerabilities’ Detection by Analysing Application Execution Traces

Abstract: Over the years, digital traces have proven to be significant for analyzing IT systems, including applications. With the persistent threats arising from the widespread proliferation of malware and the evasive techniques employed by cybercriminals, researchers and application vendors alike are concerned about finding effective solutions. In this article, we assess a hybrid approach to detecting software vulnerabilities based on analyzing traces of application execution. To accomplish this, we initially extract permissions and features from manifest files. Subsequently, we employ a tracer to extract events from each running application, utilizing a set of elements that indicate the behavior of the application. These events are then recorded in a trace. We convert these traces into features that can be utilized by machine learning algorithms. Finally, to identify vulnerable applications, we train these features using six machine learning algorithms (KNN, Random Forest, SVM, Naive Bayes, Decision Tree-CART, and MLP). The selection of these algorithms is based on the outcomes of several preliminary experiments. Our results indicate that the SVM algorithm produces the best performance, followed by Random Forest, achieving an accuracy of 98%for malware detection and 96% for benign applications. These findings demonstrate the relevance and utility of analyzing real application behavior through event analysis.

Author 1: Gouayon Koala

Author 2: Didier Bassol´e

Author 3: Telesphore Tiendrebeogo

Author 4: Oumarou Si´e

Keywords: Execution traces; events; vulnerability detection; malware; applications

Paper 137: Hybrid Encryption Algorithm for Information Security in Hadoop

Abstract: Network security has gained importance in recent years. Information system security is greatly aided by the development of encryption as a solution. To safeguard the shared information, several strategies are required. Thanks to the cutting-edge Internet, networking corporations, health information, and cloud applications, our data is growing exponentially every minute. In order to handle enormous amounts of data efficiently, a new application called Hadoop distributed file system (HDFS) was created. However, HDFS doesn’t come with any built-in data encryption tools, which poses serious security risks. In order to increase data security, encryption techniques are established; nevertheless, standard algorithms fall short when dealing with bigger files. In this study, huge data will be secured using a novel hybrid encryption algorithm that combines CP-ABE (encryption based on the features of the encryption policy), AES (advanced standard encryption), and RSA (Rivest-Shamir-Adleman). The suggested model’s performance is compared against that of traditional encryption algorithms like DES, 3DES, and Blowfish in order to demonstrate improved performance as it relates to decryption time, encryption time, and throughput. The results of the studies demonstrate that our suggested method’s algorithm is more secure.

Author 1: Youness Filaly

Author 2: Fatna El mendili

Author 3: Nisrine Berros

Author 4: Younes El Bouzekri EL idrissi

Keywords: Hadoop distributed file system (HDFS); big data security; data encryption; data decryption

Paper 138: Review of Unsupervised Segmentation Techniques on Long Wave Infrared Images

Abstract: This paper studies the different unsupervised segmentation algorithms that have been proposed and their efficacy on thermal images. The scope of this research is to develop a generalized approach to blindly segment urban thermal imagery to assist the system in identifying regions by shape instead of pixel values. Most methods can be classified as thresholding, edge-based, region-based, clustering, or texture analysis. We explained methods, worked before applying the methods of interest on thermal images of 8-bit and 16-bit resolution, and evaluated the performance. The evaluation section discusses where each method succeeded, where it failed, and how the performance can be enhanced. Finally, we study the time complexity of each method to assess the feasibility of implementing a fast, and generalized method of pixel labeling.

Author 1: Mohammed Abuhussein

Author 2: Aaron L. Robinson

Author 3: Iyad Almadani

Keywords: Unsupervised segmentation; thermal images; texture analysis; pixel labeling; Gabor; GMM; image analysis; K-Means; MRF; Otsu’s; DNN; region-based clustering

Paper 139: The Application of Intelligent Evaluation Method with Deep Learning in Calligraphy Teaching

Abstract: Scientific and effective teaching quality evaluation (QE) is helpful to improve teaching mode and improve teaching quality. At present, calligraphy teaching (CT) QE methods are few in number and have poor evaluation effect. Aiming at these problems, deep learning (DL) is introduced to realize intelligent evaluation of CT quality. First, based on relevant research, the CTQE indicator system is constructed. Secondly, rough set and the principal component analysis (PCA) are used to reduce the dimension of the CTQE index system and extract four common factors. Then, the corresponding index data is input into the BP neural network (BPNN) model optimized by the improved sparrow search algorithm for fitting. Finally, combining the above contents, the improved sparrow search algorithm (ISSA) BPNN model is built to realize the intelligent evaluation of CT quality. The experimental results show that the loss value of ISSA-BPN model is 0.21, and the fitting degree of CT data is 0.953. The evaluation Accuracy is 95%, Precision is 0.945, Recall is 0.923, F1 is 0.942, and AUC is 0.967. These values are superior to the most advanced teaching QE model available. The SSA-BPNNCTQE model proposed in the study has excellent performance in CTQE. This is of positive significance to the improvement of teaching quality and students' calligraphy level.

Author 1: Yu Wang

Keywords: Deep learning; calligraphy teaching; BPNN; intelligent evaluation; sparrow search algorithm

Paper 140: Deep Learning-based Mobile Robot Target Object Localization and Pose Estimation Research

Abstract: Two key technologies in robotic object grasping are target object localization and pose estimation (PE), respectively, and the addition of a robotic vision system can dramatically enhance the flexibility and accuracy of robotic object grasping. The study optimizes the classical convolutional structure in the target detection network considering the limited computing power and memory resources of the embedded platform, and replaces the original anchor frame mechanism using an adaptive anchor frame mechanism in combination with the fused depth map. For evaluating the target’s pose, the smooth plane of its surface is identified using the semantic segmentation network, and the target’s pose information is obtained by solving the normal vector of the plane, so that the robotic arm can absorb the object surface along the direction of the plane normal vector to achieve the target’s grasping. The adaptive anchor frame can maintain an average accuracy of 85.75% even when the number of anchor frames is increased, which proves its anti-interference ability to the over fitting problem. The detection accuracy of the target localization algorithm is 98.8%; the accuracy of the PE algorithm is 74.32%; the operation speed could be 25 frames/s. It could satisfy the requirements of real-time physical grasping. In view of the vision algorithm in the study, physical grasping experiments were carried on. Then the success rate of object grasping in the experiments was above 75%, which effectively verified the practicability.

Author 1: Caixia He

Author 2: Laiyun He

Keywords: Mobile robot; target object localization; pose estimation; YOLOv2 network; FCN semantic segmentation network

Paper 141: Character Representation and Application Analysis of English Language and Literature Based on Neural Network

Abstract: The development of computer technology has promoted the continuous progress of Natural language processing technology and the great development of ideology and culture, and also prompted literary workers to create a large number of literary works. This poses a new challenge to the application of Natural language processing technology. Text analysis and processing is realized by Natural language processing technology. In the information society, the amount of data is increasing exponentially, and the number of literary works produced is also rapidly increasing. In order to gain a comprehensive understanding of domestic and foreign history and culture, some Chinese readers are not only satisfied with reading Chinese works from ancient and modern times, but also hope to read and understand foreign literary works. Current mainstream methods for literary character analysis are manual, making the results highly subjective and inefficient for large-scale literary works. To address this problem, this study proposes a character representation and analysis method based on neural networks using English novels as an example. By preprocessing data and utilizing the word dependency relationship to represent character vectors and calculate similarity, the study uses the Skip-gram model to train character vectors and K-means for clustering. An AGA-BPNN model is proposed for character and gender prediction and classification, with a 95.42% accuracy rate achieved in character prediction classification, and an average accuracy, recall, and F1 score of 0.953, 0.962, and 0.962, respectively, in gender prediction and classification. The results demonstrate the effectiveness of the method and propose a new approach for novel character analysis.

Author 1: Yao Song

Keywords: Neural network; English; literary image; character vector; similarity calculation

Paper 142: NLPashto: NLP Toolkit for Low-resource Pashto Language

Abstract: In recent years, natural language processing (NLP) has transformed numerous domains, becoming a vital area of research. However, the focus of NLP studies has predomi-nantly centered on major languages like English, inadvertently neglecting low-resource languages like Pashto. Pashto, spoken by a population of over 50 million worldwide, remains largely unexplored in NLP research, lacking off-the-shelf resources and tools even for fundamental text-processing tasks. To bridge this gap, this study presents NLPashto, an open-source and publicly accessible NLP toolkit specifically designed for Pashto. The initial version of NLPashto introduces four state-of-the-art models for Spelling Correction, Word Segmentation, Part-of-Speech (POS) Tagging, and Offensive Language Detection. The toolkit also includes essential NLP resources like pre-trained static word embeddings, Word2Vec, fastText, and GloVe. Furthermore, we have pre-trained a monolingual language model for Pashto from scratch, using the Bidirectional Encoder Representations from Transformers (BERT) architecture. For the training and evaluation of all the models, we have developed several benchmark datasets and also included them in the toolkit. Experimental results demonstrate that the models exhibit satisfactory perfor-mance in their respective tasks. This study can be a significant milestone and will hopefully support and speed-up future research in the field of Pashto NLP.

Author 1: Ijazul Haq

Author 2: Weidong Qiu

Author 3: Jie Guo

Author 4: Peng Tang

Keywords: NLP; text processing; word segmentation; POS tagging; BERT, LLMs; Pashto; low-resource languages; CRF; CNNs; RNNs

Paper 143: Intelligent Traffic Video Retrieval Model based on Image Processing and Feature Extraction Algorithm

Abstract: Intelligent transportation is a system that combines data-driven information with traffic management to achieve intelligent monitoring and retrieval functions. In order to further improve the retrieval accuracy of the system model, a new retrieval model was designed. The functional requirements of the system were summarized, and the three stages of data preprocessing, feature matching, and feature extraction were analyzed in detail. The study adopted preprocessing measures such as equalization and normalization to minimize the negative effects of noise and brightness. Based on the performance of various algorithms, the distance method was selected as the feature matching method, which has a wider applicability and is better at processing bulk data. Next, the study utilizes Euclidean distance method to extract keyframes and divides the feature extraction into three parts: color, shape, and texture. The methods of color moment, canny operator, and grayscale co-occurrence matrix are used to extract them, and ultimately achieve relevant image retrieval. The research conducted multiple experiments on the retrieval performance of the model, and analyzed the results of retrieving single and mixed features. The experimental results showed that the algorithm performed better in the face of mixed feature extraction. Compared with the average value of a single feature, the recall and precision of the three mixed features increased by 13.78% and 15.64%, respectively. Moreover, in the case of a large number of concurrent features, the algorithm also met the basic requirements. When the concurrent number was 100, the average response time of the algorithm is 4.46 seconds. Therefore, the algorithm proposed by the research institute effectively improves the ability of video retrieval and can meet the requirements of timeliness, which can be widely applied in practical applications.

Author 1: Xiaoming Zhao

Author 2: Xinxin Wang

Keywords: Matching extraction; feature fusion; image retrieval; intelligent transportation

Paper 144: The Mechanism of the Role of Big Data Knowledge Management in the Development of Enterprise Innovation

Abstract: The effectiveness and efficiency of enterprise knowledge management depends on the effectiveness and efficiency of the enterprise's implementation of knowledge management. Big data technology can collect, analyse and apply the massive amount of data in an organisation to support the implementation of knowledge management. Therefore, exploring the role of big-data knowledge management in the development of enterprise innovation will help enterprises to better implement knowledge management. Based on this, the study aims to propose a model for predicting big data knowledge management and enterprise innovation development for high-tech enterprises in China. The study firstly used Principal Component Analysis (PCA) to decrease the dimensionality of the model, and then used the particle swarm algorithm to optimize BP neural network (PSO-BP). Network (PSO-BP) was used to evaluate enterprise knowledge management and enterprise innovation development. The results of the study show that the absolute values of the relative errors of the pre-processed model do not exceed the 5% threshold, and only the relative errors of some indicators are relatively large, such as X5 and X7, with values of 4.5% and -3.8%, indicating that the model has a good performance in predicting the innovation effect of enterprises.

Author 1: Guangyu Yan

Author 2: Rui Ma

Keywords: Big data knowledge management; BP neural network algorithm; enterprise innovation development; principal component analysis; particle swarm optimization algorithm; correlation analysis

Paper 145: A Roadmap Towards Optimal Resource Allocation Approaches in the Internet of Things

Abstract: Introducing new technologies has facilitated people's lives more than ever. As one of these emerging technologies, the Internet of Things (IoT) enables objects we handle daily to interact with each other or humans and exchange information through the Internet by being equipped with sensors and communication technologies. IoT turns the physical world into a virtual world where heterogeneous objects and devices can be interconnected and controlled. IoT-based networks face numerous challenges, including energy and sensor transmission limitations. New technologies are needed to spread the IoT platform, optimize costs, cover heterogeneous connections, reduce power consumption, and diminish delays. Users of IoT-based systems typically use services that are integrated into these networks. Service providers provide users with on-demand services. The interrelationship between this request and response must be managed in a way that is done using a resource allocation strategy. Therefore, resource allocation plays a major role in these systems and networks. The allocation of resources involves matters such as how much, where, and when available resources should be provided to the user economically. The allocation of resources in the IoT environment is also subject to various challenges, including maintaining the quality of service, achieving a predetermined level of service, storing power, controlling congestion, and reducing costs. As the IoT resource allocation problem is an NP-Hard one, many research efforts have been conducted on this topic, and various algorithms have been developed. This paper reviews published publications on IoT resource allocation, outlining the underlying principles, the latest developments, and current trends.

Author 1: Jiyin Zhou

Keywords: Internet of things; resource utilization; resource allocation; systematic review

Paper 146: Social Media Mining to Detect Online Violent Extremism using Machine Learning Techniques

Abstract: In this paper, we explore the challenging domain of detecting online extremism in user-generated content on social media platforms, leveraging the power of Machine Learning (ML). We employ six distinct ML and present a comparative analysis of their performance. Recognizing the diverse and complex nature of social media content, we probe how ML can discern extremist sentiments hidden in the vast sea of digital communication. Our study is unique, situated at the intersection of linguistics, computer science, and sociology, shedding light on how coded language and intricate networks of online communication contribute to the propagation of extremist ideologies. The goal is twofold: not only to perfect detection strategies, but also to increase our understanding of how extremism proliferates in digital spaces. We argue that equipping machine learning algorithms with the ability to analyze online content with high accuracy is crucial in the ongoing fight against digital extremism. In conclusion, our findings offer a new perspective on online extremism detection and contribute to the broader discourse on the responsible use of ML in society.

Author 1: Shynar Mussiraliyeva

Author 2: Kalamkas Bagitova

Author 3: Daniyar Sultan

Keywords: NLP; machine learning; social networks; extremism detection; textual contents

Paper 147: Text Mining-based Enterprise Financial Performance Evaluation in the Context of Enterprise Digital Transformation

Abstract: As enterprises gradually move towards digitalization, it is increasingly difficult to accurately evaluate changes in corporate financial performance. To improve this situation, the study uses a text mining algorithm based on the web crawler principle to extract keywords from corporate annual reports, select representative financial performance indicators through IF-FDP, and construct a corporate financial performance evaluation model using the entropy weighting method. The performance comparison experiments of the text mining algorithm proposed in the study show that the accuracy-recall rate area under the line of the text mining algorithm proposed in the study is 0.83 and the average F-value is 0.34, which are both better than other algorithms. In the empirical analysis of the financial performance evaluation model, it was found that the financial performance evaluation model had the smallest absolute error of 0.3%, which was lower than the other models. The above results indicate that both the text mining algorithm and the performance evaluation model proposed in the study outperform the comparison algorithm and model. Therefore, the performance evaluation model proposed by the study can be used to effectively evaluate the financial performance of enterprises accurately and promote the development of enterprises, which has practical application value.

Author 1: Changrong Guo

Author 2: Jing Xing

Keywords: Web crawler; text mining; IF-FDP; entropy method; financial performance; evaluation model

Paper 148: Study of the Drug-related Adverse Events with the Help of Electronic Health Records and Natural Language Processing

Abstract: Surveillance of pharmacovigilance, also known as drug safety surveillance, involves the monitoring and evaluation of drug-related adverse events or side effects to ensure the safe and effective use of medications. Pharmacovigilance is an essential component of healthcare systems worldwide and plays a crucial role in identifying and managing drug safety concerns. Natural language processing (NLP) can play a crucial role in surveillance activities within pharmacovigilance by analyzing and extracting information from various sources, such as clinical trial reports, electronic health records, social media, and scientific literature. It is important to note that while NLP can be a powerful tool in pharmacovigilance surveillance, it should always be used in conjunction with human expertise. NLP algorithms can assist in the identification and extraction of relevant information, but the final assessment and decision-making should involve the knowledge and judgment of trained pharmacovigilance professionals. In this paper, we intend to train and test our models using the dataset from the Medication, Indication, and Adverse Drug Events challenge. This dataset will include patient notes as well as entity categories such as Medication, Indication, and ADE, as well as various sorts of relationships between these entities. Because ADE-related information extraction is a two-stage process, the outcome of the second step (i.e., relation extraction) will be utilized to compare all models. The analysis of drug-related adverse events using electronic health records and automated approaches can considerably increase the effectiveness of ADE-related information extraction, although this depends on the methodology, data, and other aspects. Our findings can help with ADE detection and NLP research.

Author 1: Sarah Allabun

Author 2: Ben Othman Soufiene

Keywords: Natural language processing; surveillance of pharmacovigilance; drug-related adverse