IJACSA Volume 14 Issue 4 - thesai.org

IJACSA Volume 14 Issue 4

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: An End-to-End Deep Learning System for Recommending Healthy Recipes Based on Food Images

Abstract: Healthy food leads to healthy living and it is a major issue in our days. Nutri-Score is a nutrition label that can be calculated from the nutritional values of a food and helps evaluating the healthiness of it. Nevertheless, we don’t always have the nutritional values of the food, so it is not always easy identifying this label. In the same way, it is not easy finding the healthier option to a favorite food. In this paper an end-to end deep learning system is proposed to identify the Nutri-Score label and recommend similar but healthier recipes based on food images. A new dataset of images is extracted from the Recipe 1M and labeled with the Nutri-Score value calculated for each image. Pretrained models Resnet50, Resnet101, EfficientNetB2 and DensNet121 are tuned based on this dataset. The embeddings from the last convolutional layer of the input image are used to find its most similar neighbor based on KNN algorithm. The proposed system suggests recipes with the lowest Nutri-Score similar to the inputted image. Implementations show that the Resnet50 provides highest prediction accuracy.

Author 1: Ledion Lico

Author 2: Indrit Enesi

Author 3: Sai Jawahar Reddy Meka

Keywords: Deep learning; nutri-score; new dataset; healthy food; accuracy

Paper 2: An Automatic Framework for Number Plate Detection using OCR and Deep Learning Approach

Abstract: The use of automatic number plate detection devices in safety, commercial, and security has increased over the past few years. Number plate detection using computer vision is used to provide fast and accurate detection and recognition. Lately, many computerized approaches have been developed for the identification of vehicle registration details based on license plate numbers using either Deep Learning (DL) methodologies. In the proposed framework, we used Optical Character Recognition (OCR) and a deep learning-based new approach for automatic number plate detection and recognition. A deep learning approach trains the model to recognize the vehicle. The vehicle registration plate area is cropped adequately from the image, and a Convolution Neural Network (CNN) uses OCR to identify numbers and letters. The Jetson TX2 NVIDIA target served as the model's training data source, and its performance has been tested on a public dataset from Kaggle database. We obtained the highest accuracy of 96.23%. The proposed system could recognize vehicle license plate numbers on real-world images. The system can be implemented at security checkpoint entrances in highly restricted areas such as military areas or areas surrounding high-level government agencies.

Author 1: Yash Shambharkar

Author 2: Shailaja Salagrama

Author 3: Kanhaiya Sharma

Author 4: Om Mishra

Author 5: Deepak Parashar

Keywords: Number plat detection; recognition; deep learning; OCR; image classification

Paper 3: Convolution Neural Networks for Phishing Detection

Abstract: Phishing is one of the significant threats in cyber security. Phishing is a form of social engineering that uses e-mails with malicious websites to solicitate personal information. Phishing e-mails are growing in alarming number. In this paper we propose a novel machine learning approach to classify phishing websites using Convolution Neural Networks (CNNs) that use URL based features. CNNs consist of a stack of convolution, pooling layers, and a fully connected layer. CNNs accept images as input and perform feature extraction and classification. Many CNN models are available today. To avoid vanishing gradient problem, recent CNNs use entropy loss function with Rectified Linear Units (ReLU). To use a CNN, we convert feature vectors into images. To evaluate our approach, we use a dataset consists of 1,353 real world URLs that were classified into three categories-legitimate, suspicious, and phishing. The images representing feature vectors are classified using a simple CNN. We developed MATLAB scripts to convert vectors into images and to implement a simple CNN model. The classification accuracy obtained was 86.5 percent.

Author 1: Arun D. Kulkarni

Keywords: Classification; convolution neural networks; machine learning; phishing URLs

Paper 4: A Radial Basis Network-based Early Warning Algorithm for Physical Injuries in Marathon Athletes

Abstract: For marathon runners, a single injury may affect their lifelong athletic career, so their injury management is very important. The current injury management for marathon runners has a certain lag, and the current injury warning is mainly based on manual teams, which is costly and poorly automated. To solve these problems, the study proposes a marathon athlete physical injury warning algorithm based on inertia weight adjustment optimized radial basis network. Particle swarm optimization technology has also been incorporated into early warning algorithms. Finally, an athlete injury and disease early warning model is constructed based on the algorithm. The results of performance tests show that the algorithm has a minimum fitness function value of 0.13, which is significantly lower than the current algorithm used for comparison. In the test with real data, the MAPE of the proposed algorithm was as low as 7.598% and the agreement of the hazard score results with the expert human assessment reached 100%. The results of the study indicate the practicality of the algorithm to assist work teams and perform early warning of physical injuries in athletes. However, the high number of iterations required is a limitation awaiting resolution.

Author 1: Ruisheng Jiao

Author 2: Juan Luo

Keywords: Radial basis neural network; exponentially decreasing inertia weights; early warning algorithm; sports injury; marathon; particle swarm; model building

Paper 5: Classification of Hand Movements Based on EMG Signals using Topological Features

Abstract: Hand movement classification based on Electromyo-graphy (EMG) signals has been extensively investigated in the past decades as a promising approach used for controlling upper prosthetics or robotics. Topological data analysis is a relatively new and increasingly popular tool in data science that uses mathematical techniques from topology to analyze and understand complex data sets. This paper proposes a method for classifying hand movements based on EMG signals using topological features crafted with the tools of TDA. The main findings of this work on hand movement EMG classification are as follows: (1) topological features are effective in classifying EMG signals and outperform other time domain features tested in the experiments; (2) the 0-th Betti numbers are more effective than the 1-st Betti numbers; (3) Betti amplitude is a more stable and powerful feature than other topological features discussed in this paper. Additionally, Betti curves were used to visualize topological patterns for hand movement EMG.

Author 1: Jianyang Li

Author 2: Lei Yang

Author 3: Yunan He

Author 4: Osamu Fukuda

Keywords: EMG classification; persistent homology; topological features; betti curve

Paper 6: Research on Automatic Intrusion Detection Method of Software-Defined Security Services in Cloud Environment

Abstract: In a cloud environment, software defined security services are highly vulnerable to malicious virus attacks. In response to software security issues, this project plans to use machine learning technology to achieve automated detection of software security services in a cloud environment. Firstly, study the intrusion characteristics of software defined security services in cloud environments based on piecewise sample regression, and establish their statistical feature quantities. Then, using the method of decision statistical analysis, achieve its fixed identification. Finally, the intrusion characteristics of software defined security services in the cloud environment are studied and compared with the data in the cloud environment to obtain its power spectral density. On this basis, machine learning methods are used to extract features from software security services in the cloud environment, in order to achieve the goal of automatic extraction and optimization of software security services in the cloud environment. Through simulation experiments, the credibility of the proposed algorithm for software defined security services in the cloud environment was verified, and the attack characteristics of software defined security services in the cloud environment were effectively patched.

Author 1: Xingjie Huang

Author 2: Jing Li

Author 3: Jinmeng Zhao

Author 4: Beibei Su

Author 5: Zixian Dong

Author 6: Jing Zhang

Keywords: Cloud environment; software; security services; invasion; detection; machine learning

Paper 7: Experimental Analysis of WebHDFS API Throughput

Abstract: Data analysis is very important for the success of any business today. It helps to optimize business processes, analyze users’ behavior, demands etc. There are powerful data analytics tools, such as the ones of the Hadoop ecosystem, but they require multiple high-performance servers to run and high-qualified experts to install, configure and support them. In most cases, small companies and start-ups could not afford such expenses. However, they can use them as web services, on demand, and pay much lower fees per request. To do that, companies should somehow share their data with an existing, already deployed, Hadoop cluster. The most common way of uploading their files to the Hadoop’s Distributed File System (HDFS) is through the WebHDFS API (Application Programming Interface) that allows remote access to HDFS. For that reason, the API’s throughput is very important for the efficient integration of a company’s data to the Hadoop cluster. This paper performs a series of experimental analyses aiming to determine the WebHDFS API’s throughput, if it is a bottleneck in integration of a company’s data to existing Hadoop infrastructure and to detect all possible factors that influence the speed of data transmission between the clients’ software and the Hadoop’ file system.

Author 1: Yordan Kalmukov

Author 2: Milko Marinov

Keywords: WebHDFS API; throughput analysis; data analytical tools; Hadoop Distributed File System (HDFS)

Paper 8: Gradually Generative Adversarial Networks Method for Imbalanced Datasets

Abstract: Imbalanced dataset can cause obstacles to classification and result in a decrease in classification performance. There are several methods that can be used to deal the data imbalances, such as methods based on SMOTE and Generative Adversarial Networks (GAN). These methods are used for overcoming data oversampling so that the amount of minority data can increase and it can reach a balance with the majority data. In this research, the selected dataset is classified as a small imbalanced dataset of less than 200 records. The proposed method is the Gradually Generative Adversarial Network (GradGAN) model which aims to handle data imbalances gradually. The stages of the GradGAN model are adding the original minority dataset gradually so that it will create new minority datasets until a balance of data is created. Based on the algorithm flow described, the minority data is multiplied by the value of the variable that has been determined repeatedly to produce new balanced minority data. The test results on the classification of datasets from the GradGAN model produce an accuracy value of 8,3% when compare to that without GradGAN.

Author 1: Muhammad Misdram

Author 2: Muljono

Author 3: Purwanto

Author 4: Edi Noersasongko

Keywords: Classification; imbalance; GAN model; GradGAN model; significant oversampling

Paper 9: Human Fall Detection for Smart Home Caring using Yolo Networks

Abstract: In order to help the elderly and limit the incidence of falls that result in injuries, effective fall detection in smart home applications is a challenging topic. Many techniques have been created employing both vision and non-vision-based technologies. Many researchers have been drawn to the vision-based technique amongst them because of its viability and application. However, there is still room for improvement in the effectiveness of fall detection given the poor accuracy rate and high computational cost issues with current vision-based techniques. This study introduces a new dataset for posture and fall detection, whose photo images were gathered from Internet resources and data augmentation. It employs YOLO networks for fall detection purpose. Furthermore, different YOLO networks are implemented on our dataset to address the most accurate and effective model. Based on assessment parameters including accuracy, F1 score, recall, and mAP, the performance of the various YOLOv5n, s and YOLOv6s versions are compared. As experimental results showed, the YOLOv5s performed better than other.

Author 1: Bo LUO

Keywords: YOLO; computer vision; fall detection; smart home; caring

Paper 10: Investigation of You Only Look Once Networks for Vision-based Small Object Detection

Abstract: Small object detection is a challenging issue in computer vision-based algorithms. Although various methods have been investigated for common objects including person, car and others, small object are not addressed in this issue. Therefore, it is necessary to conduct more researches on them. This paper is focused on small object detection especially jewellery as current object detection methods suffer from low accuracy in this domain. This paper introduces a new dataset whose images were taken by a web camera from a jewellery store and data augmentation procedure. It comprises three classes, namely, ring, earrings, and pendant. In view of the small target of jewellery and the real-time detection, this study adopted the You Only Look Once (Yolo) algorithms. Different Yolo based model including eight versions are implemented and train them using our dataset to address most effective one. Evaluation criteria, including accuracy, F1 score, recall, and mAP, are used to evaluate the performance of the various YOLOv5, YOLOv6, and YOLOv7 versions. According to the experimental findings, utilizing YOLOv6 is significantly superior to YOLOv7 and marginally superior to YOLOv5.

Author 1: Li YANG

Keywords: YOLOv7; YOLOv6; YOLOv5; computer vision; jewellery detection; small object detection; real-time detection

Paper 11: A Novel Data Aggregation Method for Underwater Wireless Sensor Networks using Ant Colony Optimization Algorithm

Abstract: Underwater Wireless Sensor Networks (UWSNs) have a wide range of applications for monitoring the ocean and exploring the offshore environment. Sensor nodes are typically dispersed throughout the area of interest at different depths in these networks. Sensor nodes on the seabed must use a routing protocol in order to communicate with surface-level nodes. The suitability assessment considers network resources, application requirements, and environmental factors. By combining these factors, a platform for resource-aware routing strategies can be created that meet the needs of different applications in dynamic environments. Numerous challenges and problems are associated with UWSNs, including the lack of battery power, instability of topologies, a limited bandwidth, long propagation times, and interference from the ocean. These problems can be addressed through the design of routing protocols. The routing protocol facilitates the transfer of data between source and destination nodes. Data aggregation and UWSN protocols are widely used to achieve better outcomes. This paper describes an energy-aware algorithm for data aggregation in UWSNs that uses the improved ACO (Ant Colony Optimization) algorithm to maximize the packet delivery ratio, improve the network lifetime, decrease end-to-end delay, and use less energy.

Author 1: Lianchao Zhang

Author 2: Jianwei Qi

Author 3: Hao Wu

Keywords: UWSNs; routing; data aggregation; energy efficiency; ant colony optimization algorithm

Paper 12: A New Machine Learning-based Hybrid Intrusion Detection System and Intelligent Routing Algorithm for MPLS Network

Abstract: Machine Learning (ML) is seen as a promising application that offers autonomous learning and provides optimized solutions to complex problems. The current Multiprotocol Label Switching (MPLS)-based communication system is packed with exponentially increasing applications and different Quality-of-Services (QoS) requirements. As the network is getting complex and congested, it will become challenging to satisfy the QoS requirements in the MPLS network. This study proposes a hybrid ML-based intrusion detection system (ML-IDS) and ML-based intelligent routing algorithm (ML-RA) for MPLS network. The research is divided into three parts, which are (1) dataset development, (2) algorithm development, and (3) algorithm performance evaluation. The dataset development for both algorithms is carried out via simulations in Graphical Network Simulator 3 (GNS3). The datasets are then fed into MATLAB to train ML classifiers and regression models to classify the incoming traffic as normal or attack and predict traffic delays for all available routes, respectively. Only the normal traffic predicted by the ML-IDS algorithm will be allowed to enter the network domain, and the route with the fastest delay predicted by the ML-RA is assigned for routing. The ML-based routing algorithm is compared to the conventional routing algorithm, Routing Information Protocol version 2 (RIPv2). From the performance evaluations, the ML-RA shows 100 percent accuracy in predicting the fastest route in the network. During network congestion, the proposed ML outperforms the RIPv2 in terms of delay and throughput on average by 57.61 percent and 46.57 percent, respectively.

Author 1: Mohammad Azmi Ridwan

Author 2: Nurul Asyikin Mohamed Radzi

Author 3: Kaiyisah Hanis Mohd Azmi

Author 4: Fairuz Abdullah

Author 5: Wan Siti Halimatul Munirah Wan Ahmad

Keywords: Machine learning; intrusion detection system; routing algorithm; quality of service; communication system

Paper 13: Plant Disease Classification and Adversarial Attack based CL-CondenseNetV2 and WT-MI-FGSM

Abstract: In recent years, deep learning has been increasingly used to the detection of pests and diseases. Unfortunately, deep neural networks are particularly vulnerable when attacked by adversarial examples. Hence it is vital to explore the creation of intensely aggressive adversarial examples to increase neural network robustness. This paper proposes a wavelet transform and histogram equalization-based adversarial attack algorithm: WT-MI-FGSM. In order to verify the performance of the WT-MI-FGSM, we propose a plant pests and diseases identification method based on the coordinate attention mechanism and CondenseNetV2: CL-CondenseNetV2. The accuracy of CL- CondenseNetV2 on the PlantVillage dataset is 99.45%, which indicates that the improved CondenseNetV2 model has a more significant classification performance. In adversarial sample experiments using WT-MI-FGSM and CL-CondenseNetV2, experimental results show that when CL-CondenseNetV2 is attacked by the adversarial algorithm WT-MI-FGSM, the error rate reaches 89.8%, with a higher attack success rate than existing adversarial attack algorithms. In addition, the accuracy of CL-CondenseNetV2 is improved to 99.71% by adding the adversarial samples generated by WT-MI-FGSM to the training set and performing adversarial training. The experiments demonstrate that the adversarial examples caused by WT-MI-FGSM can improve the model's performance.

Author 1: Yong Li

Author 2: Yufang Lu

Keywords: Adversarial examples; FGSM; plants diseases and pests; attention mechanism; CondenseNetV2

Paper 14: EMOCASH: An Intelligent Virtual-Agent Based Multiplayer Online Serious Game for Promoting Money and Emotion Recognition Skills in Egyptian Children with Autism

Abstract: Autism, often known as "autism spectrum disorders (ASD)," is one of the most common developmental disabilities that affect how people learn, behave, communicate, and interact with others. Two crucial everyday tasks that people with ASD typically struggle with are managing finances and recognizing emotions. As the online gaming sector grows and develops, the question of why this type of media can't be used as a useful educational tool for those with ASD arises. This paper discusses this issue via a novel virtual agent-based multiplayer online serious game referred to as "EMOCASH," which aims to improve these important tasks for Egyptian children with ASD and achieve transfer of acquired knowledge to real-world situations via a 3D virtual shop scenario that was designed using the Autism ASPECTSSTM Design Index. EMOCASH served as an instrument for investigating the following research question: What role does technology play in the education of those with ASD? Numerous sub-questions that were related to the primary question were also addressed. A variety of usability metrics were used to assess effectiveness, efficiency and satisfaction aspects.

Author 1: Hussein Karam Hussein Abd El-Sattar

Keywords: Autism; virtual agents; serious games; digital technology; AI; online gaming; usability and accessibility

Paper 15: Adaptive Balance Optimizer: A New Adaptive Metaheuristic and its Application in Solving Optimization Problem in Finance

Abstract: Adaptability becomes important in developing metaheuristic algorithms, especially in tackling stagnation. Unfortunately, almost all metaheuristics are not equipped with an adaptive approach that makes them change their strategy when stagnation happens during iteration. Based on this consideration, a new metaheuristic, called an adaptive balance optimizer (ABO), is proposed in this paper. ABO's unique strategy focuses on exploitation when improvement happens and switching to exploration during stagnation. ABO also uses a balanced strategy between exploration and exploitation by performing two sequential searches, whatever circumstance it faces. These sequential searches consist of one guided search and one random search. Moreover, ABO also deploys both a strict acceptance approach and a non-strict acceptance approach. In this work, ABO is challenged to solve a set of 23 classic functions as a theoretical optimization problem and a portfolio optimization problem as the use case for the practical optimization problem. In portfolio optimization, ABO should optimize the quantity of ten stocks in the energy and mining sector listed in the IDX30 index. In this evaluation, ABO is competed with five other metaheuristics: marine predator algorithm (MPA), golden search optimizer (GSO), slime mold algorithm (SMA), northern goshawk optimizer (NGO), and zebra optimization algorithm (ZOA). The simulation result shows that ABO is better than MPA, GSO, SMA, NGO, and ZOA in solving 21, 18, 16, 11, and 8, respectively, in solving 23 functions. Meanwhile, ABO becomes the third-best performer in solving the portfolio optimization problem.

Author 1: Purba Daru Kusuma

Author 2: Ashri Dinimaharawati

Keywords: Optimization; metaheuristic; adaptability; portfolio optimization; IDX30

Paper 16: Improved Speaker Recognition for Degraded Human Voice using Modified-MFCC and LPC with CNN

Abstract: Economical speaker recognition solution from degraded human voice signal is still a challenge. This article is covering results of an experiment which targets to improve feature extraction method for effective speaker identification from degraded human audio signal with the help of data science. Every speaker’s audio has identical characteristics. Human ears can easily identify these different audio characteristics and classify speaker from speaker’s audio. Mel-Frequency Cepstral Coefficient (MFCC) supports to get same intelligence in machine also. MFCC is extensively used for human voice feature extraction. In our experiment we have effectively used MFCC and Linear Predictive Coding (LPC) for better speaker recognition accuracy. MFCC first outlines frames and then finds cepstral coefficient for each frame. MFCC use human audio signal and convert it in numerical value of audio features, which is used to recognize speaker efficiently by Artificial Intelligence (AI) based speaker recognition system. This article covers how effectively audio features can be extracted from degraded human voice signal. In our experiment we have observed improved Equal Error Rate (EER) and True Match Rate (TMR) due to high sampling rate and low frequency range for mel-scale triangular filter. This article also covers pre-emphasis effects on speaker recognition when high background noise comes with audio signal.

Author 1: Amit Moondra

Author 2: Poonam Chahal

Keywords: Data science; artificial intelligence; MFCC; LPC; CNN; mel-spectrum; speaker recognition

Paper 17: Customer Segmentation of Personal Credit using Recency, Frequency, Monetary (RFM) and K-means on Financial Industry

Abstract: This research focuses on how to build a segmentation model for credit customers to identify the potential for defaulting credit customers based on their transaction history. Currently, there is no segmentation available for this possibility of payment failure. Credit scoring helps in minimizing credit risk when applying for credit. However, using RFM (Recency, Frequency, Monetary) models helps to score each transaction variable of the customer's financial activity. K-means then assists in the process of segmenting the results of the RFM model scoring, which occurs in the middle of the customer's repayment schedule. Challenge is how to decide the variable that can be used in RFM models and how to interpret the clusters that have been formed and the actual implementation of the customer. The Bank can divide the clusters that have possibility of payment failure by their customers so that banks can take preventive actions and as information for the collection system to be able to make payment withdrawals or billing.

Author 1: Hafidh Rizkyanto

Author 2: Ford Lumban Gaol

Keywords: Credit; credit risk; recency; frequency; monetary; K-means

Paper 18: Challenges of Digital Twin Technologies Integration in Modular Construction: A Case from a Manufacturer’s Perspective

Abstract: Automated models of physical objects are known as Digital twins; letting one to outline, test, array, and monitor and manage robotics in the real world. .CPS (Cyber-physical system) data have to be assembled through real life procedures to form a real-time monitoring cyber model in order to produce Digital Twin. Modification in the cyber model will be shown in real life system to guess or manage. As a result of digital and the progression in ICT, manufacturing and aviation or aerospace industries are now utilizing digital twin. Nonetheless, the uses of DT's in several production firms have not been researched massively. Those studies were sparse in construction, where structures are built without Superstructures, which thus sparked global concern. Herein, DT applications in building/manufacturing sector and in various firms were reviewed first and thereafter aimed on publications concerning DT applications in industry area by organizing a systematic search via Scopus. Notably, the publications were singled out immediately after the assessment of the publications and the study continues to investigate and evaluate the Potentials of digital twins in MIC, Restriction of digital twins in MIC, Impact of digital twins on industry, and Cost of time which should be appropriate for model development. The analysis report however demonstrated that DT is dedicated and thoughtful of in midst of inclusion along other digital technologies. More so, theoretical structure is formed in order to apply DT in module installation in MiC around the circumstances of Hong Kong which happens to be usual city case of high-density. Interestingly, the implementation of Digital twin in Modular Integrated Construction is expected to provide promising potential with significant benefits, such as improved logistics and manufacturing management by employing Digital Twins to track on-site progress during module installation.

Author 1: Laith Jamal Aldabbas

Keywords: Digital Twin; DTs; enabling technologies; digital twin model; applications; challenges; literature review

Paper 19: SuffixAligner: A Python-based Aligner for Long Noisy Reads

Abstract: Third-generation sequencing technologies have revolutionized genomics research by generating long reads that resolve many computational challenges such as long genomics variations and repeats. Mapping a set of sequencing reads against a reference genome is the first step of many genomic data analysis pipelines. Many mapping/alignment tools are introduced and always made different compromises between the alignment accuracy and the resource usage in terms of memory space and processor speed. SuffixAligner is a python-based aligner for long noisy reads generated from third-generation sequencing machines. SuffixAligner follows the seed extending approach and exploits the nature of the biological alphabet that has a fixed size and a predefined lexical ordering to construct a suffix array for indexing a reference genome. A suffix array is used to efficiently search the indexed reference and locate the exactly matched seeds among the reads and the reference. The matched seeds are arranged into windows/clusters and the ones with the maximum number of seeds are reported as candidates for mapping positions. Using real data sets from third-generation sequencing experiments, we evaluated SuffixAligner against lordFAST, BWA, GEM3, and Minimap2, in which the results showed that SuffixAligner mapped more reads compared to the other compared tools. The source code of SuffixAligner is available at: https://github.com/ZeinabRabea/SuffixAligner.

Author 1: Zeinab Rabea

Author 2: Sara El-Metwally

Author 3: Samir Elmougy

Author 4: M. Z. Rashad

Keywords: Long reads sequencing; reads mapping; suffix array; alignment; seed extending; LF mapping

Paper 20: Scrum: A Systematic Literature Review

Abstract: This study presents a Systematic Literature Review on an agile project management tool. The study offers a brief comparison between traditional and agile project management methodologies. Their respective concepts and characteristics are laid out to highlight and explain their main differences. The agile methods include quantitative and qualitative data, showing Scrum framework characteristics. This study highlights the importance of project management in function of its emergence as a response to problems encountered during improperly conducted projects. Furthermore, this study provides relevant information for professionals in the Industrial Engineering area and computer science. The results allowed us to conclude that Scrum is an agile framework for empirical-based project development; it was developed in the 1990s by Jeff Sutherland. It is a flexible and adaptable methodology. Scrum research peaked in 2020, and continues to be studied, mainly in the field of computer science. Finally, Brazil is well-positioned in third place for works published.

Author 1: Adrielle Cristina Sassa

Author 2: Isabela Alves de Almeida

Author 3: Tábata Nakagomi Fernandes Pereira

Author 4: Milena Silva de Oliveira

Keywords: Project management; agile methods; agile manifesto; scrum

Paper 21: Artificial Intelligence Based Modelling for Predicting CO2 Emission for Climate Change Mitigation in Saudi Arabia

Abstract: Climate change (such as global warming) causes the barrier in the attaining sustainable development goals. Emission of greenhouse gases (primarily carbon dioxide CO2 emission) are the root cause of global warming. This research analyses and investigates the emission of CO2 and attempts to develop an optimal model to forecast the CO2 emission. Several machine learning and statistical modeling techniques have been implemented and evaluated to explore the patterns and trends of CO2 emissions to develop an optimal model for forecasting future CO2 emissions. The implemented methods include such as Exponential Smoothing, Transformers, Temporal Convolutional Network (TCN), and neural basis expansion analysis for interpretable time series. The data for training these models have been collected and synthesized from various sources using a web crawler. The performance of these models has been evaluated using various performance measurement metrices such as RMSE, R2 score, MAE, MAPE and OPE. The N-BEATS model demonstrated an overall better performance for forecasting CO2 emission in Saudi Arabia in comparison to the other models. In addition, this paper also provides recommendations and strategies for mitigating the climate change (by reducing CO2 emission).

Author 1: Sultan Alamri

Author 2: Shahnawaz Khan

Keywords: Exponential smoothing; transformers; temporal convolutional network; neural basis expansion analysis; climate change

Paper 22: Implementation of Revised Heuristic Knowledge in Average-based Interval for Fuzzy Time Series Forecasting of Tuberculosis Cases in Sabah

Abstract: Fuzzy time series forecasting is one method used to forecast in certain reality problems. The research on fuzzy time series forecasting has been increased due to its capability in dealing with vagueness and uncertainty. In this paper, we are dealing with implementation of revised heuristic knowledge to basic average-based interval and showing that these models forecast better than the basic one. We suggest three different lengths of interval, size 5, size 10 and size 20 to be used in comparing these models of average-based interval, average-based interval with implementation of heuristic knowledge and, average-based interval with implementation of revised heuristic knowledge. These models applied to forecast the number of tuberculosis cases reported monthly in Sabah starting from January 2012 until December 2019. A few numerical examples are shown as well. The performances of evaluations are shown by comparison on the values obtained by Mean Square error (MSE) and Root Mean Square Error (RMSE).

Author 1: Suriana Lasaraiya

Author 2: Suzelawati Zenian

Author 3: Risman Mat Hasim

Author 4: Azmirul Ashaari

Keywords: Fuzzy time series; forecasting; length of interval; average-based interval; heuristic knowledge

Paper 23: Event Feature Pre-training Model Based on Public Opinion Evolution

Abstract: The comments in the evolution of network public opinion events not only reflect the attitude of netizens towards the event itself, but also are the key basis for mastering the dynamics of public opinion. According to the comment data in the event evolution process, an event feature vector pre-training model NL2ER-Transformer is constructed to realize the real-time automatic extraction of event features. Firstly, a semi-supervised multi-label curriculum learning model is proposed to generate comment words, event word vectors, event words, and event sentences, so that a public opinion event is mapped into a sequence similar to vectorized natural language. Secondly, based on the Transformer structure, a training method is proposed to simulate the evolution process of events, so that the event vector generation model can learn the evolution law and the characteristics of reversal events. Finally, the event vectors generated by the presented NL2ER-Transformer model are compared with the event vectors generated by the current mainstream models such as XLNet and RoBerta. This paper tests the pre-trained model NL2ER-Transformer and three pre-trained benchmark models on four downstream classification models. The experimental results show that using the vectors generated by NL2ER-Transformer to train downstream models compared to using the vectors generated by other pre-trained benchmark models to train downstream models, the accuracy, recall, and F1 values are 16.66%, 44.44%, and 19% higher than the best downstream model. At the same time, in the evolutionary capability analysis test, only four events show partial errors. In terms of performance of semi-supervised model, the proposed semi-supervised multi-label curriculum learning model outperforms mainstream models in four indicators by 6%, 23%, 8%, and 15%, respectively.

Author 1: WANG Nan

Author 2: TAN Shu-Ru

Author 3: XIE Xiao-Lan

Author 4: LI Hai-Rong

Author 5: JIANG Jia-Hui

Keywords: Event vectorization; NL2ER-transformer model; public opinion reversal prediction; evolution of public opinion event; multi label semi supervised learning

Paper 24: A Novel Network Intrusion Detection System Based on Semi-Supervised Approach for IoT

Abstract: An intrusion detection system (IDS) is one of the most effective ways to secure a network and prevent unauthorized access and security attacks. But due to the lack of adequately labeled network traffic data, researchers have proposed several feature representations models over the past three years. However, these models do not account for feature generalization errors when learning semantic similarity from the data distribution and may degrade the performance of the predictive IDS model. In order to improve the capabilities of IDS in the era of Big Data, there is a constant need to extract the most important features from large-scale and balanced network traffic data. This paper proposes a semi-supervised IDS model that leverages the power of untrained autoencoders to learn latent feature representations from a distribution of input data samples. Further, distance function-based clustering is used to find more compact code vectors to capture the semantic similarity between learned feature sets to minimize reconstruction loss. The proposed scheme provides an optimal feature vector and reduces the dimensionality of features, reducing memory requirements significantly. Multiple test cases on the IoT dataset MQTTIOT2020 are conducted to demonstrate the potential of the proposed model. Supervised machine learning classifiers are implemented using a proposed feature representation mechanism and are compared with shallow classifiers. Finally, the comparative evaluation confirms the efficacy of the proposed model with low false positive rates, indicating that the proposed feature representation scheme positively impacts IDS performance.

Author 1: Durga Bhavani A

Author 2: Neha Mangla

Keywords: IoT; security; intrusion detection system; semi-supervised; autoencoder; clustering; machine learning

Paper 25: Iris Recognition Through Edge Detection Methods: Application in Flight Simulator User Identification

Abstract: To meet the increasing security requirement of authorized users of flight simulators, personal identification is becoming more and more important. Iris recognition stands out as one of the most accurate biometric methods in use today. Iris recognition is done through different edge detection methods. Therefore, it is important to have an understanding of the different edge detection methods that are in use these days. Specifically, the biomedical research shows that irises are as different as fingerprints or the other patterns of the recognition. Furthermore, because the iris is a visible organism, its exterior look can be examined remotely using a machine vision system. The main part of this paper delves into concerns concerning the selection of the best results giving method of the recognition. In this paper, three edge detection methods, namely Canny, Sobel and Prewitt, are applied to the image of eye (iris) and their comparative analysis is discussed. These methods are applied using the Software MATLAB. The datasets used for this purpose are CASIA and MMU. The results indicate that the performance of Canny edge detection method is best as compared to Sobel and Prewitt. Image quality is a key requirement in image-based object recognition. This paper provides the quality evaluation of the images using different metrics like PSNR, SNR, MSE and SSIM. However, SSIM is considered best image quality metric as compared to PSNR, SNR and MSE.

Author 1: Sundas Naqeeb Khan

Author 2: Samra Urooj Khan

Author 3: Onyeka Josephine Nwobodo

Author 4: Krzysztof Adam. Cyran

Keywords: Identification; authentication; detection; canny; Sobel; Prewitt; PSNR; SNR; SSIM; MSE

Paper 26: Personalized Music Recommendation Based on Interest and Emotion: A Comparison of Multiple Algorithms

Abstract: Recommendation algorithms can greatly improve the efficiency of information retrieval for users. This article briefly introduced recommendation algorithms based on association rules and algorithms based on interest and emotion analysis. After crawling music and comment data from the NetEase Cloud platform, a simulation experiment was conducted. Firstly, the performance of the Back-Propagation Neural Network (BPNN) in the interest and emotion-based algorithm for recommending music was tested, and then the impact of the proportion of emotion weight between comments and music on the emotion analysis-based algorithm was tested. Finally, the three recommendation algorithms based on association rules, user ratings, and interest and emotion analysis were compared. The results showed that when the BPNN used the dominant interest and emotion and secondary interest and emotion as judgment criteria, the accuracy of interest and emotion recognition for music and comments was higher. When the proportion of interest and emotion weight between comments and music was 6:4, the interest and emotion analysis-based recommendation algorithm had the highest accuracy. The interest and emotion-based recommendation algorithm had higher recommendation accuracy than the association rule-based and user rating-based algorithms, and could provide users with more personalized and emotional music recommendations.

Author 1: Xiuli Yan

Keywords: Interest and emotion; recommendation algorithm; music; personalization

Paper 27: Research on Customer Retention Prediction Model of VOD Platform Based on Machine Learning

Abstract: Advanced wireless technology and smart mobile devices allow users to watch Internet video from almost anywhere. The major VOD platforms are competing with each other for customers, slowly shifting from a "product-centric" strategic goal to a "customer-centric" one. At present, existing research is limited to platform business model and development strategy as well as user behavior research, but there is less research on customer retention prediction. In order to effectively solve the customer retention prediction problem, this study applies machine learning methods to video-on-demand platform customer retention prediction, improves the traditional RFM model to establish the RFLH theoretical model for video-on-demand platform customer retention prediction, and uses machine learning methods to predict the number of customer retention days. The Optuna algorithm is used to determine the model hyperparameters, and the SHAP framework is integrated to analyze the important factors affecting customer retention. The experimental results show that the comprehensive performance of the LightGBM model is better than other models. The total number of user logins in the past week, the length of video playback in the same day, and the time difference between the last login and the present are important features that affect customer retention prediction. This study can help companies develop effective customer management strategies to maximize potential customer acquisition and existing customer retention for maximum market advantage.

Author 1: Quansheng Zhao

Author 2: Zhijie Zhao

Author 3: Liu Yang

Author 4: Lan Hong

Author 5: Wu Han

Keywords: Video-on-demand platform; Customer Retention Forecast; RFM Model; Machine Learning; SHAP

Paper 28: Development of Computer Vision-enabled Augmented Reality Games to Increase Motivation for Sports

Abstract: This research paper presents the development of computer vision-enabled augmented reality games based on action detection to increase motivation for sports. With the increasing popularity of digital games, physical activity and sports participation have been declining, especially among the younger generation. To address this issue, we developed a series of augmented reality games that require players to perform physical actions to progress and succeed in the game. These games were developed using computer vision technology to detect the players' movements and provide real-time feedback, enhancing the gaming experience and promoting physical activity. The results of our user study showed that participants who played the augmented reality games reported higher levels of motivation to engage in physical activity and sports. The findings suggest that computer vision-enabled augmented reality games can be an effective tool to promote physical activity and sports participation, especially among younger generations.

Author 1: Bauyrzhan Doskarayev

Author 2: Nurlan Omarov

Author 3: Bakhytzhan Omarov

Author 4: Zhuldyz Ismagulova

Author 5: Zhadra Kozhamkulova

Author 6: Elmira Nurlybaeva

Author 7: Galiya Kasimova

Keywords: Augmented reality; computer vision; action detection; action classification; machine learning

Paper 29: Opposition Learning Based Improved Bee Colony Optimization (OLIBCO) Algorithm for Data Clustering

Abstract: Clustering of data in case of data mining has a major role in recent research as well as data engineers. It supports for classification and regression type of problems. It needs to obtain the optimized clusters for such application. The partitional clustering and meta-heuristic search techniques are two helpful tools for this task. However the convergence rate is one of the important factors at the time of optimization. In this paper, authors have taken a data clustering approach with improved bee colony algorithm and opposition based learning to improve the rate of convergence and quality of clustering. It introduces the opposite bees that are created using opposition based learning to achieve better exploration. These opposite bees occupy exactly the opposite position that of the mainstream bees in the solution space. Both the mainstream and opposite bees explore the solution space together with the help of Bee Colony Optimization based clustering algorithm. This boosts the explorative power of the algorithm and hence the convergence rate. The algorithm uses a steady state selection procedure as a tool for exploration. The crossover and mutation operation is used to get balanced exploitations. This enables the algorithm to avoid sticking in local optima. To justify the effectiveness of the algorithm it is verified with the open datasets from the UCI machine learning repository as the benchmark. The simulation result shows that it performs better than some benchmark as well as recently proposed algorithms in terms of convergence rate, clustering quality, and exploration and exploitation capability.

Author 1: Srikanta Kumar Sahoo

Author 2: Priyabrata Pattanaik

Author 3: Mihir Narayan Mohanty

Author 4: Dilip Kumar Mishra

Keywords: Bee colony optimization; BCO based clustering; data clustering; partitional clustering; meta-heuristic search

Paper 30: Fuzzy Rank-Based Ensemble Model for Accurate Diagnosis of Osteoporosis in Knee Radiographs

Abstract: The main factor in fractures among seniors and women post-menopausal is osteoporosis, which decreases the density of bones. Finding a low-cost diagnostic technology to identify osteoporosis in its initial stages is imperative considering the substantial expenses of diagnosis and therapy. The simplest and most widely used imaging method for detecting bone diseases is X-ray radiography, however, it is problematic to manually examine X-rays for osteoporosis as well as to identify the essential components and choose elevated classifiers. To categorize x-ray pictures of knee joints into normal, osteopenia, and osteoporosis condition categories, authors present a process in this investigation that uses three convolutional neural networks (CNN) architectures, i.e., Inception v3, Xception, and ResNet 18, to create an ensemble-based classifier model. The suggested ensemble approach employs a fuzzy rank-based unification of classifiers by taking into account two distinct parameters on the decision scores produced by the aforementioned base classifiers. Contrary to the straightforward fusion strategies that have been mentioned in the literature, the suggested ensemble methodology finalizes predictions on the test specimens by considering the confidence in the recommendations of the base learners. A 5-fold cross-validation approach has been employed to assess the developed framework using a benchmark dataset that has been made accessible to the general population. The suggested model yields an accuracy rate of 93.5% with a loss of 0.082. Further, the AUC is observed to be 98.1, 97.9 and 97.3 for normal, osteopenia and osteoporosis, respectively. The results demonstrate the model’s usefulness by outperforming various state-of-the-art approaches.

Author 1: Saumya Kumar

Author 2: Puneet Goswami

Author 3: Shivani Batra

Keywords: Convolutional Neural Network; diagnosis; knee; osteoporosis; transfer learning models; X-rays

Paper 31: A Novel Deep CNN-RNN Approach for Real-time Impulsive Sound Detection to Detect Dangerous Events

Abstract: In this research paper, we presented a novel approach to detect impulsive sounds in real-time using a combination of Deep CNN and RNN architectures. The proposed approach was evaluated using our collected dataset of impulsive sounds, and the results showed that it outperformed traditional audio signal processing methods in terms of accuracy and F1-score. The proposed approach has several advantages over traditional methods, including the ability to handle complex audio patterns, detect impulsive sounds in real-time, and improve its performance with a large dataset of labeled impulsive sounds. However, there are some limitations to the proposed approach, including the requirement for a large amount of labeled data to train effectively, environmental factors that may impact the accuracy of the detection, and high computational requirements. Overall, the proposed approach demonstrates the effectiveness of using a combination of Deep CNN and RNN architectures for impulsive sound detection, with potential applications in various fields such as public safety, industrial settings, and home security systems. The proposed approach is a significant step towards developing automated systems for detecting dangerous events and improving public safety.

Author 1: Nurzhigit Smailov

Author 2: Zhandos Dosbayev

Author 3: Nurzhan Omarov

Author 4: Bibigul Sadykova

Author 5: Maigul Zhekambayeva

Author 6: Dusmat Zhamangarin

Author 7: Assem Ayapbergenova

Keywords: CNN; RNN; deep learning; impulsive sound; dangerous sound; artificial intelligence

Paper 32: Reversible De-identification of Specific Regions in Biomedical Images and Secured Storage by Randomized Joint Encryption

Abstract: In many circumstances, de-identification of a specific region of a biomedical image is necessary. De-identification is used to hide the subject’s identity or to prevent the display of the objectionable or offensive region(s) of the image. The concerned region can be blurred (de-identified) by using a suitable image processing technique guided by the region-defining mask. The proposed method provides lossless blurring, which means the original image can be recovered fully with zero loss. The blurred image and the region-defining mask, along with the digital signature, are jointly encrypted to form the composite cipher matrix, and it is stored in the cloud for further distribution. The composite cipher matrix is decrypted to recover the blurred image by the conventional end user. Further, using the deblur key, the original image can be recovered with zero loss by the fully authorized special end users. On decryption, the digital signature is available for both types of end users. The proposed method uses randomized joint encryption using integer matrix keys in a finite field. The experimental results show that the proposed method achieves a reduction in the average execution time of encryption by 30 to 40 percent compared to its nearest competitor. Additionally, the proposed scheme achieves very nearly ideal performance with reference to the correlation coefficient, entropy, pixel change rate, and structural similarity index. Overall, the proposed algorithm performs substantially better than the other similar existing schemes for large-sized images.

Author 1: Prabhavathi K

Author 2: Anandaraju M. B

Keywords: Region identification mask; modular matrix inverse; selective image encryption; image de-identification; randomized joint encryption; image authentication

Paper 33: Texture Analytics for Accurate Person Recognition: A Multimodal Approach

Abstract: Securing the resources is a most challenging task in the digital era. Traditionally, password and ID card systems were used to provide security. Password and ID cards can be stolen or hacked; to overcome this drawback biometric systems are used to authenticate the user to access the data or resources. Biometric system uses physical and behavioral characteristics of the user. Biological characteristics of the person like face, fingerprint, iris, palm print, voice, hand geometry etc. cannot be stolen and misused. Even though unimodal biometric system is more secure as compared to the traditional approach, it is not able to handle intra-class, inter-class variations, noisy data and spoofing attack. These problems can be solved using multimodal biometrics. In this paper, we discuss unimodal biometric system using Local Binary Pattern (LBP) and Local Ternary Pattern (LTP). We propose a feature level fusion of face and fingerprint biometric traits using LTP. The implementation of the introduced system stands in comparison to the unimodal LBP and LTP for face and fingerprint system. The system is tested on ORL, UMIST, VISA face dataset and FVC fingerprint dataset. Experimental results show that the multimodal biometric system using LTP gives better accuracy as compared to the unimodal biometric system.

Author 1: Suchetha N V

Author 2: Sharmila Kumari M

Keywords: Unimodal; Multi-modal; LBP; LTP; intra-class; inter-class; spoofing attack

Paper 34: Deep Learning Models for Crime Intention Detection Using Object Detection

Abstract: The majority of visual based surveillance applications and security systems heavily rely on object detection, which serves as a critical module. In the context of crime scene analysis, images and videos play an essential role in capturing visual documentation of a particular scene. By detecting objects associated with a specific crime, police officers are able to reconstruct a scene for subsequent analysis. Nevertheless, the task of identifying objects of interest can be highly arduous for law enforcement agencies, mainly because of the massive amount of data that must be processed. Hence, the main objective of this paper is to propose a DL-based model for detecting tracked objects such as handheld firearms and informing the authority about the threat before the incident happens. We have applied VGG-19, ResNet, and GoogleNet as our deep learning models. The experiment result shows that ResNet50 has achieved the highest average accuracy of 0.92% compared to VGG19 and GoogleNet, which have achieved 0.91% and 0.89%, respectively. Also, YOLOv6 has achieved the highest MAP and inference speed compared to the faster R-CNN.

Author 1: Abdirahman Osman Hashi

Author 2: Abdullahi Ahmed Abdirahman

Author 3: Mohamed Abdirahman Elmi

Author 4: Octavio Ernest Romo Rodriguez

Keywords: Object detection; deep learning; crime scenes; video surveillance; convolutional neural network; YOLOv6

Paper 35: Hand Gesture Recognition Based on Various Deep Learning YOLO Models

Abstract: Some varieties of sign languages are used by deaf or hard-of-hearing people worldwide to interact with others more effectively, consequently sign language's automatic translation is expressive and important. Significant improvements in computer vision have been made recently, notably in tasks based on object detection using deep learning. By locating things in visual photos or videos, the genuine cutting-edge one-step object detection approach greatly provides exceptional detection accuracy. With the help of messaging or video calling, this study suggests a technique to get beyond these obstacles and enhance communication for such persons, regardless of their disability. To recognize motions and classes, we provide an enhanced model based on Yolo (You Look Only Once) V3, V4, V4-tiny, and V5. The dataset is clustered using the suggested algorithm, requiring only manual annotation of a reduced number of classes and analysis for patterns that aid in target prediction. The suggested method outperforms the current object detection approaches based on the YOLO model, according to experimental results.

Author 1: Soukaina Chraa Mesbahi

Author 2: Mohamed Adnane Mahraz

Author 3: Jamal Riffi

Author 4: Hamid Tairi

Keywords: Neural network; deep learning; YOLO; object detection; hand gesture

Paper 36: Optimized Image Authentication Algorithm using Redundant Wavelet Transform Based Sift Descriptors and Complex Zernike Moments

Abstract: Due to the advanced multimedia editing tools and supported by sophisticated hardware, creating image/video manipulations for malicious purposes is increasing which is almost impossible to detect manually. Moreover, to conceal the traces, different post-processing operations are performed. Therefore, authenticity is a growing concern and important for identifying original and forged images. One of the popular image manipulations is copy-move forgery in which one or more regions in the image are duplicated to create a malicious effect within an image. The work in this article presents redundant wavelet transform based complex Zernike moment and Scale Invariant Feature Transform (SIFT) keypoint matching technique for copy-move image forgery operation detection. SIFT is robust against scale and rotation that works on identifying similarity using exhaustive search of SIFT features. After extracting SIFT keypoint features agglomerative hierarchical clustering is employed for grouping and key point matching operation is performed. Finally, block matching operation is evaluated and forged regions in the manipulated images are marked and displayed. This work also presents optimized SIFT key-point feature computations resulting in lower computation time, often one of the requirements in real time deployment. The proposed algorithm performance is evaluated using Precision, Recall, F-Measure and average detection accuracy on popular and publicly available MICC-220 database. The proposed technique demonstrates improved speed-up and detection rate compared to existing approaches.

Author 1: Pooja Vijayakumaran Kallath

Author 2: Kondaka Lakshmisudha

Keywords: Forgery detection; scale invariant feature transform; key point operation; block matching; agglomerative hierarchical clustering

Paper 37: Anchor-free Proposal Generation Network for Efficient Object Detection

Abstract: Deep learning object detection methods are usually based on anchor-free or anchor-based scheme for extracting object proposals and one-stage or two-stage structure for producing final predictions. As each scheme or structure has its own strength and weakness, combining their strength in a unified framework is an interesting research topic. However, this topic has not attracted much attention in recent years. This paper presents a two-stage object detection method that utilizes an anchor-free scheme for generating object proposals in the initial stage. For proposal generation, this paper employs an efficient anchor-free network for predicting object corners and assigns object proposals based on detected corners. For object prediction, an efficient detection network is designed to enhance both detection accuracy and speed. The detection network includes a lightweight binary classification subnetwork for removing most false positive object candidates and a light-head detection subnetwork for generating final predictions. Experimental results on the MS-COCO dataset demonstrate that the proposed method outperforms both anchor-free and two-stage object detection baselines in terms of detection performance.

Author 1: Hoanh Nguyen

Keywords: Object detection; deep learning; convolutional neural network; proposal generation network

Paper 38: Detecting Fraud Transaction using Ripper Algorithm Combines with Ensemble Learning Model

Abstract: In the context of the 4.0 technology revolution, which develops and applies strongly in many fields, in which the banking sector is considered to be the leading one, the application of algorithms to detect fraud is extremely important. necessary. In recent years, credit card transactions including physical credit card payments and online payments have become increasingly popular in many countries around the world. This convenient payment method attracts more and more criminals, especially credit card fraud. As a result, many banks around the world have developed fraud detection and prevention systems for each credit card transaction. Data mining is one of the techniques applied in these systems. This study uses the Ripper algorithm to detect fraudulent transactions on large data sets, and the results obtained with accuracy, recall, and F1 measure of more than 97%. This research then used the Ripper algorithm combined with Ensemble Learning models to detect fraudulent transactions, the results are more than 99% reliable. Specifically, this model using the Ripper algorithm combined with the Gradient Boosting method has improved the predictive ability and obtained very reliable results. The use of algorithms combined with machine learning models is expected to be a new topic and will be widely applied to banks’ or organizations’ activities related to e-commerce.

Author 1: Vo Hoang Khang

Author 2: Cao Tung Anh

Author 3: Nguyen Dinh Thuan

Keywords: Financial fraud; data mining; credit card fraud; transaction; ensemble methods

Paper 39: A Single-valued Pentagonal Neutrosophic Geometric Programming Approach to Optimize Decision Maker’s Satisfaction Level

Abstract: Achieving the desired level of satisfaction for a decision-maker in any decision-making scenario is considered a challenging endeavor because minor modifications in the process might lead to incorrect findings and inaccurate decisions. In order to maximize the decision-maker’s satisfaction, this paper proposes a Single-valued Neutrosophic Geometric Programming model based on pentagonal fuzzy numbers. The decision-maker is typically assumed to be certain of the parameters, but in reality, this is not the case, hence the parameters are presented as neutrosophic fuzzy values. The decision-maker, with this strategy, is able to achieve varying levels of satisfaction and dissatisfaction for each constraint and even complete satisfaction for certain constraints. Here the decision maker aims to achieve the maximum level of satisfaction while maintaining the level of hesitation and minimizing dissatisfaction in order to retain an optimum solution. Furthermore, transforming the objective function into a constraint adds one more layer to the N-dimensional multi-parametrizes α ,β and γ. The advantages of this multi-parametrized proposed method over the existing ones are proven using numerical examples.

Author 1: Satyabrata Nath

Author 2: Purnendu Das

Author 3: Pradip Debnath

Keywords: Decision making; pentagonal neutrosophic numbers; single-valued neutrosophic geometric programming; multi-parametric programming

Paper 40: Multifaceted Sentiment Detection System (MSDS) to Avoid Dropout in Virtual Learning Environment using Multi-class Classifiers

Abstract: Sentiment analysis with machine learning plays a vital role in Higher Educational Institutions (HEI) for decision making. Technology-enabled interactions can only be successful when a strong student-teacher link is established, and the emotions of students are clearly comprehended. The paper aims at proposing Multifaceted Sentiment Detection System (MSDS) for detecting sentiments of higher education students participating in virtual learning and to classify the comments posted by them using Machine Learning (ML) algorithms. Present research evaluated a total of n=1590 students’ comments with the presence of three specific multifaceted characteristics each providing 530 comments to perform Sentiment Analysis (SA) for monitoring their sentiments, opinions that facilitate predicting dropout in virtual learning environment (VLE). This begins with the phrase extraction; then data pre-processing techniques namely digits, punctuation marks and stop-words removal, spelling correction, tokenization, lemmatization, n-grams, and POS (Part of Speech) are applied. Texts are vectorized using two feature extraction techniques with count vectorization and TF-IDF metrics and classified with four multiclass supervised ML techniques namely Random Forest, Linear SVC, Multinomial Naive Bayes, and Logistic Regression for multifaceted sentiment classification. Analyzing students’ feedback using sentiment analysis techniques classifies their positive, negative, or even more refined emotions that enables dropout prediction. Experimental results reveal that the highest mean accuracy result for device efficiency, cognitive behavior, technological expertise with cloud learning platform usage were achieved by Logistic Regression with 98.49%, Linear SVC with 93.58% and Linear SVC with 92.08% respectively. Practically, results confirm feasibility for detecting students’ multifaceted behavioral patterns and risk of dropout in VLE.

Author 1: Ananthi Claral Mary. T

Author 2: Arul Leena Rose. P. J

Keywords: Sentiment analysis; opinions; TF-IDF; n-gram; virtual learning; machine learning; NLTK; text pre-processing

Paper 41: Implementation of CNN for Plant Identification using UAV Imagery

Abstract: Plants are the world's most significant resource since they are the only natural source of oxygen. Additionally, plants are considered crucial since they are the major source of energy for humanity and have nutritional, therapeutic, and other benefits. Image identification has become more prominent in this technology-driven world, where many innovations are happening in this sphere. Image processing techniques are increasingly being used by researchers to identify plants. The capacity of Convolutional Neural Networks (CNN) to transfer weights learned with huge standard datasets to tasks with smaller collections or more particular data has improved over time. Several applications are made for image identification using deep learning, and Machine Learning (ML) algorithms. Plant image identification is a prominent part of such. The plant image dataset of about 300 images collected by mobile phone and camera from different places in the natural scenes with nine species of different plants are deployed for training. A five-layered convolution neural network (CNN) is applied for large-scale plant classification in a natural environment. The proposed work claims a higher accuracy in plant identification based on experimental data. The model achieves the utmost recognition rate of 96% NU108 dataset and UAV images of NU101 have achieved an accuracy of 97.8%.

Author 1: Mohd Anul Haq

Author 2: Ahsan Ahmed

Author 3: Jayadev Gyani

Keywords: Convolutional Neural Networks (CNN); Machine Learning (ML) algorithms; plant image identification; plant image dataset

Paper 42: An IoT-based Framework for Detecting Heart Conditions using Machine Learning

Abstract: A lot of diseases may be preventable if they can be analyzed or predicted from patient historical and family data. Predicting diagnosis depends on the gathered clinical and physiological data of patients. The more collected clinical and medical healthcare data, the more knowledge the medical support system may support. Hence, real monitoring clinical and healthcare data for patients is the trend of this decade based on Internet of Things technologies (IoT). IoT models facilitate human life by easily collecting clinical data remotely for recognizing diseases that are easily treatable if it is diagnosed early. This paper proposes a framework consisting of two models: (i) heart attack detection model (HADM); (ii) Electrocardiosignal ECG heartbeat multiclass-classification model (ECG-HMCM). Gridsearch is used to the hyperparameters optimization for different machine learning (ML) techniques. The used dataset in HADM consists of 1190 patients and 14 features. As the foundation of diagnosing cardiovascular disease is arrhythmia detection hence, we propose an ECG heartbeat multi-class classification model using MIT-BIH Arrhythmia and PTB Diagnostic ECG signals dataset which contains five categories with 109446 samples. K Nearest Neighbor (KNN) technique is applied to build ECG-HMCM in addition to the using of Gridsearch algorithm for hyperparameter optimization aiming to improve the accuracy of classification which achieved 97.5%. The proposed framework aims to facilitate human life by easily collecting clinical data remotely. The outcomes of the experiments show that the suggested framework works well in a practical setting.

Author 1: Mona Alnaggar

Author 2: Mohamed Handosa

Author 3: T. Medhat

Author 4: M. Z. Rashad

Keywords: Medical healthcare; medical support system; real monitoring; Internet of Things (IoT); ECG signals; Gridsearch; hyperparameter optimization

Paper 43: Solar Energy Forecasting Based on Complex Valued Auto-encoder and Recurrent Neural Network

Abstract: Renewable energy is becoming a trusted power source. Energy forecasting is an important research field, which is used to provide information about the future power generation of renewable energy plants. Energy forecasting helps to safely manage the power grid by minimizing the operational cost of energy production. Recent advances in energy forecasting based on deep learning techniques have shown great success but the achieved results still too far from the target results. Ordinary deep learning models have been used for time series processing. In this paper, a complex-valued autoencoder was coupled with an LSTM neural network for solar energy forecasting. The complex-valued autoencoder was used to process the time series with the advantage of processing more complex data with more input arguments. The energy value was used as a real value and the weather condition was considered as the imaginary value. Taking into account the weather condition helps to better predict power generation. The proposed approach was evaluated on the Fingrid open data dataset. The mean absolute error (MAE), root-mean-square error (RMSE) and mean absolute percentage error (MAPE) was used to evaluate the performance of the proposed method. A comparison study was performed to prove the efficiency of the proposed approach. Reported results have shown the efficiency of the proposed approach.

Author 1: Aymen Rhouma

Author 2: Yahia Said

Keywords: Solar energy forecasting; artificial intelligence; complex-valued autoencoder; long-short term memory; deep learning

Paper 44: Classification with K-Nearest Neighbors Algorithm: Comparative Analysis between the Manual and Automatic Methods for K-Selection

Abstract: Machine learning and the algorithms it uses have been the subject of many and varied studies with the development of artificial intelligence in recent years. One of the popular and widely used classification algorithms is the nearest neighbors’ algorithm and in particular k nearest neighbors. This algorithm has three important steps: calculation of distances; selection of the number of neighbors; and the classification itself. The choice of the value for the k parameter determines the number of neighbors and is important and has a significant impact on the degree of efficiency of the created model. This article describes a study of the influence of the way the k parameter is chosen - manually or automatically. Data sets, used for the study, are selected to be as close as possible in their features to the data generated and used by small businesses - heterogeneous, unbalanced, with relatively small volumes and small training sets. From the obtained results, it can be concluded that the automatic determination of the value of k can give results close to the optimal ones. Deviations are observed in the accuracy rate and the behavior of well-known KNN modifications with increasing neighborhood size for some of the training data sets tested, but one cannot expect that the same model's parameter values (e.g. for k) will be optimally applicable on all data sets.

Author 1: Tsvetelina Mladenova

Author 2: Irena Valova

Keywords: Machine Learning; KNN; classification

Paper 45: The Impact of Design-level Class Decomposition on the Software Maintainability

Abstract: The quality of the software's internal structure tends to decay due to the adaptation to environmental changes. Therefore, it is beneficial to maintain the internal structure of the software to benefit future phases of the software life cycle. A common correlation exists between decaying internal structures and problems like software smell and maintenance costs. Refactoring is a process to maintain the internal structure of software artifacts based on the smell. Decomposition of classes is one of the most common refactoring actions based on Blob smell performed at the source code level. Moving the class decomposition process to the design artifact seems to affect the quality and maintainability of the source code positively. Therefore, studying the impact of design-level class decomposition on source code quality and software maintainability is essential to ascertain the benefits of implementing design-level class decomposition. The metrics-based evaluation shows that the design-level class decomposition positively impacts the source code quality and maintainability with the rank biserial value is 0.69.

Author 1: Bayu Priyambadha

Author 2: Tetsuro Katayama

Keywords: Design level refactoring; class decomposition; class diagram decomposition; software quality; software internal quality; software maintainability

Paper 46: Listening to the Voice of People with Vision Impairment

Abstract: Extensive research developed assistive technologies (ATs) to improve mobility for people with vision impairment (PVI). However, a limited number of PVI rely on ATs for mobility. One of the factors contributing to the limited reliability and low acceptance of ATs is the developers’ failure to consider PVI mobility traits from the target group’s perspective. Many developers and researchers proposed solutions based on their knowledge and experiences, where PVI have been excluded from several studies except for limited involvement in testing phases. Accordingly, this study aims to bridge this gap by providing comprehensive information on PVIs’ behaviors, challenges, and requirements for safe and independent outdoor mobility. Therefore, a total of 15 participants with vision impairment were involved in semi-structured interviews and two observation sessions. One key finding highlights the need for AT that complements the conventional cane and overcomes its limitations, not substituting the cane. Moreover, the proposed AT should address instant mobility and future needs simultaneously. Overall, the study contributes to providing comprehensive knowledge on PVI safe and independent mobility traits, which assist AT developers to explore the potential barriers and facilitators of the adoption of ATs among PVIs and leads to develop effective and reliable ATs that meet their needs. For future work, the researchers will develop an AT that complements the conventional cane, supports instant mobility, and enhances cognitive map formation.

Author 1: Abeer Malkawi

Author 2: Azrina Kamaruddin

Author 3: Alfian Abdul Halin

Author 4: Novia Admodisastro

Keywords: People with vision impairment; assistive technology; outdoor mobility; behaviors; challenges; requirements

Paper 47: A Deep Learning based Approach for Recognition of Arabic Sign Language Letters

Abstract: No one can deny that the deaf-mute community has communication problems in daily life. Advances in artificial intelligence over the past few years have broken through this communication barrier. The principal objective of this work is creating an Arabic Sign Language Recognition system (ArSLR) for recognizing Arabic letters. The ArSLR system is developed using our image pre-processing method to extract the exact position of the hand and we proposed architecture of the Deep Convolutional Neural Network (CNN) using depth data. The goal is to make it easier for people who have hearing problems to interact with normal people. Based on user input, our method will detect and recognize hand-sign letters of the Arabic alphabet automatically. The suggested model is anticipated to deliver encouraging results in the recognition of Arabic sign language with an accuracy score of 97,07%. We conducted a comparison study in order to evaluate proposed system, the obtained results demonstrated that this method is able to recognize static signs with greater accuracy than the accuracy obtained by similar other studies on the same dataset used.

Author 1: Boutaina Hdioud

Author 2: Mohammed El Haj Tirari

Keywords: Deep learning; hand landmark model; convolutional neural network; Arabic sign language recognition

Paper 48: Improving QoS in Internet of Vehicles Integrating Swarm Intelligence Guided Topology Adaptive Routing and Service Differentiated Flow Control

Abstract: Internet of Vehicles (IoV) is an evolution of vehicular adhoc network with concepts of internet of things (IOT). Each vehicle in IOV is an intelligent object with various capabilities like sensors, computation, storage, control etc. Vehicles can connect to any other entity in the network using various services like DSRC, C2C-CC etc. Ensuring QoS for vehicle to everything (V2X) communication is a major challenge in IoV. This work applies an integration of hybrid metaheuristics guided routing and service differentiated flow control to ensure QoS in Internet of Vehicles. Clustering based network topology is adopted with clustering based on hybrid metaheuristics integrating particle swarm optimization with firefly algorithm. Over the established clusters routing decision is done using swarm intelligence. Packet flows in the network are service differentiated and flow control is done at cluster heads to reduce the congestion in the network. High congestion in routes is mitigated with back up path satisfying the QoS constraints. Due to optimization in clustering, routing and data forwarding process, the proposed solution is able to achieve higher QoS. Through simulation analysis, the proposed solution is able to achieve 2% higher packet delivery ratio and 9.67% lower end to end packet latency compared to existing works.

Author 1: Tanuja Kayarga

Author 2: Ananda Kumar S

Keywords: Particle swarm optimization (PSO); QoS; bio inspired technique

Paper 49: A Comparative Performance Evaluation of Routing Protocols for Mobile Ad-hoc Networks

Abstract: Mobile Ad Hoc Network (MANET) is a group of wireless mobile nodes that can connect with each other over a number of hops without the need for centralized management or pre-existing infrastructure. MANET has been used in several commercial areas such as intelligent shipping systems, ad hoc gaming, and clever agriculture, and non-commercial areas such as army applications, disaster rescue, and wildlife observing domains. One of the main challenges in MANET is routing mobility management which affects the performance of MANET seriously. The routing protocols have been functionally classified into proactive routing protocols, reactive routing protocols, and hybrid routing protocols. The objective of this paper is to create observations about the advantages and disadvantages of these protocols. Thus, the aim of this paper is to conduct a comparative analysis of the three groups of MANET routing protocols by comparing their features and methods in terms of routing overhead, scalability, delay, and other factors. It was shown that the proactive protocols guarantee the availability of the routes. However, it suffers from scalability and overhead. Whereas, reactive protocols initiate route discovery only when data needs to be sent. However, reactive protocols introduce an undesirable delay due to route establishment, which affects the network performance. Hybrid protocols, attempt to utilize the beneficial features of both reactive and proactive protocols, hybrid protocols are suitable for large networks and keep up-to-date information, but they increase operational complexity. It was concluded that MANET needs enhancement with regard to routing in order to meet the required performance.

Author 1: Baidaa Hamza Khudayer

Author 2: Lial Raja Alzabin

Author 3: Mohammed Anbar

Author 4: Ragad M Tawafak

Author 5: Tat-Chee Wan

Author 6: Abir AlSideiri

Author 7: Sohail Iqbal Malik

Author 8: Taief Alaa Al-Amiedy

Keywords: MANET; routing protocols; proactive protocols; reactive protocols; hybrid protocols; Ad Hoc Networks

Paper 50: Deep Learning for Combined Water Quality Testing and Crop Recommendation

Abstract: The field of agriculture and its specifics has been gaining more attention nowadays due to the limited present resources and the continuously increasing need for food. In fact, agriculture has benefited greatly from the advancements of artificial intelligence, namely, Machine Learning (ML). In order to make the most of a crop field, one must initially plan on what crop is best for planting in this particular field, and whether it will provide the necessary yield. Additionally, it’s very important to constantly monitor the quality of soil and water for irrigation of the selected crop. In this paper, we are going to rely on Machine Learning and data analysis to decide the type of crop that we will use, and the quality of soil and water. To do so, certain parameters must be taken into consideration. For choosing the crop, parameters such as sun exposure, humidity, soil pH, and soil moisture will be taken into consideration. On the other hand, water pH, electric conductivity, content of minerals such as chloride, calcium, and magnesium are among the parameters taken into consideration for water quality classification. After acquiring datasets for crop and water potability, we implemented a deep learning model in order to predict these two features. Upon training, our neural network model achieved 97% accuracy for crop recommendation and 96% for water quality prediction. However, the SVM model achieves 96% for crop recommendation and 92% for water quality prediction.

Author 1: Tahani Alkhudaydi

Author 2: Maram Qasem Albalawi

Author 3: Jamelah Sanad Alanazi

Author 4: Wejdan Al-Anazi

Author 5: Rahaf Mansour Alfarshouti

Keywords: Deep learning; irrigation; artificial intelligence; soil moisture

Paper 51: Context Aware Automatic Subjective and Objective Question Generation using Fast Text to Text Transfer Learning

Abstract: Online learning has gained a tremendous popularity in the last decade due to the facility to learn anytime, anything, anywhere from the ocean of web resources available. Especially the lockdown all over the world due to the Covid-19 pandemic has brought an enormous attention towards the online learning for value addition and skills development not only for the school/college students, but also to the working professionals. This massive growth in online learning has made the task of assessment very tedious and demands training, experience and resources. Automatic Question generation (AQG) techniques have been introduced to resolve this problem by deriving a question bank from the text documents. However, the performance of conventional AQG techniques is subject to the availability of large labelled training dataset. The requirement of deep linguistic knowledge for the generation of heuristic and hand-crafted rules to transform declarative sentence into interrogative sentence makes the problem further complicated. This paper presents a transfer learning-based text to text transformation model to generate the subjective and objective questions automatically from the text document. The proposed AQG model utilizes the Text-to-Text-Transfer-Transformer (T5) which reframes natural language processing tasks into a unified text-to-text-format and augments it with word sense disambiguation (WSD), ConceptNet and domain adaptation framework to improve the meaningfulness of the questions. Fast T5 library with beam-search decoding algorithm has been used here to reduce the model size and increase the speed of the model through quantization of the whole model by Open Neural Network Exchange (ONNX) framework. The keywords extraction in the proposed framework is performed using the Multipartite graphs to enhance the context awareness. The qualitative and quantitative performance of the proposed AQG model is evaluated through a comprehensive experimental analysis over the publicly available Squad dataset.

Author 1: Arpit Agrawal

Author 2: Pragya Shukla

Keywords: Automatic question generation; Text-to-Text-transfer-transformer (T5); natural language processing; word sense disambiguation (WSD); domain adaptation; multipartite graphs; beam-search decoding

Paper 52: Multistage End-to-End Driver Drowsiness Alerting System

Abstract: Drowsiness in drivers is the major cause for these fatal road accidents. Hence detecting drowsiness in drivers and alerting them on time is very important to avoid accidents. Researchers have developed several techniques to detect drowsiness in driver and warn the driver. However, in the past there is no work done towards the end-to-end driver drowsiness alerting system. Therefore, in this proposed system, it will ensure that the driver is awake through its end-to-end multi-stage (i.e., three stage) alerting system. The proposed system, at first performs driver authentication. Next, it detects the driver’s face and also checks whether he/she has consumed alcohol or not, in either case the car engine will not start, and a warning mail is sent. Then the system performs drowsiness detection. If the driver is found drowsy then a multi-stage alerting system (i.e., voice alert, seat vibration alert and physical alert) is performed to wake him/her. After the voice alert, the driver has to give his/her fingerprint as proof for not being drowsy. If the system fails to get a fingerprint it starts the vibration alert. Once again system asks for driver’s fingerprint, without which the system starts physical alert through robot arm which is performed with three different frequencies (i.e., Low, Medium and High) and three questions are asked after each frequency to make sure the driver is alert. In the process, it creates a log file which contains the driver’s drowsiness details, after analyzing which it gives rating to the driver and mail this rating to the concerned person. This rating can be used to choose the driver for a safe and comfortable journey. Thus, the system ensures that driver is alert and avoids road accidents.

Author 1: Sowmyashree P

Author 2: Sangeetha J

Keywords: Driver drowsiness detection; internet of things; voice alert; seat-vibration alert; physical alert; driver rating; haar cascade classifier; eye aspect ratio

Paper 53: Insights on Data Security Schemes and Authentication Adopted in Safeguarding Social Network

Abstract: With the increased social network usage, there is a rising concern about potential security and privacy risks related to digital information data. Although there have been numerous studies in this area, a summary is necessary to understand the effectiveness of existing security approaches. The ultimate goal is to provide valuable insights into the effectiveness of existing security schemes in the social network ecosystem. Therefore, the proposed paper discusses the existing research that has been done on authentication and data security measures, including methodologies, issues, benefits, and drawbacks. The inquiry further contributes to highlighting existing research trends and identifying the gap. The paper concludes by stating its learning results that help to open possible insights into the effectiveness of existing security schemes in the social network. Furthermore, blockchain is witnessed with increased interest in distributed security over large data. The paper's outcome states that blockchain-based authentication systems possess better scope if subjected to amending their inherent shortcomings. The findings of this paper emphasize the importance of continuous innovation in data security to ensure the safety and privacy of user data in an ever-evolving digital landscape. This paper offers a foundational aspect for future research toward developing more secure, privacy-preserving solutions for social network users.

Author 1: Nithya S

Author 2: Rekha B

Keywords: Social network; security threat; authentication; blockchain

Paper 54: A Novel Framework for Semi-supervised Multiple-label Image Classification using Multi-stage CNN and Visual Attention Mechanism

Abstract: To train deep neural networks effectively, a lot of labeled data is typically needed. However, real-time applications make it difficult and expensive to acquire high-quality labels for the data because it takes skill and knowledge to accurately annotate multiple label images. In order to enhance classification performance, it is also crucial to extract image features from all potential objects of various sizes as well as the relationships between labels of numerous label images. The current approaches fall short in their ability to map the label dependencies and effectively classify the labels. They also perform poor to label the unlabeled images when small amount of labeled images available for classification. In order to solve these issues, we suggest a new framework for semi-supervised multiple object label classification using multi-stage Convolutional neural networks with visual attention (MSCNN)and GCN for label co-occurrence embedding(LCE) (MSCNN-LCE-MIC), which combines GCN and attention mechanism to concurrently capture local and global label dependencies throughout the entire image classification process. Four main modules make up MSCNN-LCE-MIC: (1) improved multi-label propagation method for labeling largely available unlabeled image; (2) a feature extraction module using multi-stage CNN with visual attention mechanism that focuses on the connections between labels and target regions to extract accurate features from each input image; (3) a label co-existence learning that applies GCN to discover the associations between different items to create embeddings of label co-occurrence; and (4) an integrated multi-modal fusion module. Numerous tests on MS-COCO and PASCAL VOC2007 show that MSCNN-LCE-MIC significantly improves classification efficiency on mAP 84.3% and 95.8% respectively when compared to the most recent existing methods.

Author 1: Joseph James S

Author 2: Lakshmi C

Keywords: Semi-supervised; visual attention; multi-label; image classification; label propagation

Paper 55: Comparison of Predictive Machine Learning Models to Predict the Level of Adaptability of Students in Online Education

Abstract: With the onset of the COVID-19 pandemic, online education has become one of the most important options available to students around the world. Although online education has been widely accepted in recent years, the sudden shift from face-to-face education has resulted in several obstacles for students. This paper, aims to predict the level of adaptability that students have towards online education by using predictive machine learning (ML) models such as Random Forest (RF), K-Nearest-Neighbor (KNN), Support vector machine (SVM), Logistic Regression (LR) and XGBClassifier (XGB).The dataset used in this paper was obtained from Kaggle, which is composed of a population of 1205 high school to college students. Various stages in data analysis have been performed, including data understanding and cleaning, exploratory analysis, training, testing, and validation. Multiple parameters, such as accuracy, specificity, sensitivity, F1 count and precision, have been used to evaluate the performance of each model. The results have shown that all five models can provide optimal results in terms of prediction. For example, the RF and XGB models presented the best performance with an accuracy rate of 92%, outperforming the other models. In consequence, it is suggested to use these two models RF and XGB for prediction of students' adaptability level in online education due to their higher prediction efficiency. Also, KNN, SVM and LR models, achieved a performance of 85%, 76%, 67%, respectively. In conclusion, the results show that the RF and XGB models have a clear advantage in achieving higher prediction accuracy. These results are in line with other similar works that used ML techniques to predict adaptability levels.

Author 1: Orlando Iparraguirre-Villanueva

Author 2: Carmen Torres-Ceclén

Author 3: Andrés Epifanía-Huerta

Author 4: Gloria Castro-Leon

Author 5: Melquiades Melgarejo-Graciano

Author 6: Joselyn Zapata-Paulini

Author 7: Michael Cabanillas-Carbonell

Keywords: Machine learning; adaptability; students; online education; prediction; models

Paper 56: Analyzing WhisperGate and BlackCat Malware: Methodology and Threat Perspective

Abstract: The increasing use of powerful evasive ransomware malware in cyber warfare and targeted attacks is a persistent and growing challenge for nations, corporations, and small and medium-sized enterprises. This threat is evidenced by the emergence of the WhisperGate malware in cyber warfare, which targets organizations in Ukraine to render targeted devices inoperable, and the BlackCat malware, which targets large organizations by encrypting files. This paper outlines a practical approach to malware analysis using WhisperGate and BlackCat malware as samples. It subjects them to heuristic-based analysis techniques, including a combination of static, dynamic, hybrid, and memory analysis. Specifically, 12 tools and techniques were selected and deployed to reveal the malware’s innovative stealth and evasion capabilities. This methodology shows what techniques can be applied to analyze critical malware and differentiate samples that are variations of known threats. The paper presents currently available tools and their underlying approaches to performing automated dynamic analysis on potentially malicious software. The study thus demonstrates a practical approach to carrying out malware analysis to understand cybercriminals’ behavior, techniques, and tactics.

Author 1: Mathew Nicho

Author 2: Rajesh Yadav

Author 3: Digvijay Singh

Keywords: Malware analysis; WhisperGate; BlackCat; malware sample; ransomware

Paper 57: A Cloud Native Framework for Real-time Pricing in e-Commerce

Abstract: Real-time pricing is a form of 'dynamic pricing,' and it enables online sellers to adjust prices in real-time in response to variations in demand and competition to achieve higher revenue or improve customer satisfaction. As modern e-commerce implementations become more cloud-based, this paper proposes a cloud-native framework for a real-time pricing system. We take a requirement driven approach to come up with a modular architecture and a set of reusable components for real-time pricing. Following DSRM methodology, during the design phase, we identify and develop the theoretical foundations for key parts of the system, such as pricing models, competition and demand watchers, and other analytics components that fulfill the functional requirements of the system. At the stage of implementation, we describe how each of these components and the entire cloud application will be configured using an AWS cloud native implementation. As a framework, this work can support a variety of pricing models, demonstrating that multiple pricing models have been discussed. Other low-latency, reusable components described in this work provide the ability to react quickly to changes in demand and competition. We also provide a price-cache that decouples pricing model calculation from end-user price requests and keeps price query latency to a minimum. For a real-time system, where latency stands to be the most desired NFR, we validate the system for price-request latency (found to be a single digit of milliseconds) and market reaction latency (less than a second). Overall, our proposed framework provides a comprehensive solution for real-time pricing, which can be adapted to different business needs and can help online sellers optimize their pricing strategies.

Author 1: Archana Kumari

Author 2: Mohan Kumar. S

Keywords: Real-time pricing; cloud-native design; system-design; pricing-framework; Amazon web services

Paper 58: A Deep Learning Approach for Sentiment Classification of COVID-19 Vaccination Tweets

Abstract: Now-a-days, social media platforms enable people to continuously express their opinions and thoughts about different topics. Monitoring and analyzing the sentiments of people is essential for governments and business organizations to better understand people’s feelings and thoughts. The Coronavirus disease 2019 (COVID-19) has been one of the most trending topics on social media over the last two years. Consequently, one of the preventative measures to control and prevent the spread of the virus was vaccination. A dataset was formed by collecting tweets from Twitter for over a month from November 13th to December 31st, 2021. After data cleaning, the tweets were assigned a positive, negative, or neutral label using a natural language processing (NLP) sentiment analysis tool. This study aims to analyze people's public opinion towards the vaccination process against COVID-19. To fulfil this goal, an ensemble model based on deep learning (LSTM-2BiGRU) is proposed that combines long short-term memory (LSTM) and bidirectional gated recurrent unit (BiGRU). The performance of the proposed model is compared to five traditional machine learning models, two deep learning models in addition to state-of-the-art models. By comparing the results of the models used in this study, the results reveal that the proposed model outperforms all the machine and deep learning models employed in this work with a 92.46% accuracy score. This study also shows that the number of tweets that involve neutral, positive, and negative sentiments is 517496 (37%) tweets, 484258 (34%) tweets, and 409570 (29%) tweets, respectively. The findings indicate that the number of people carrying neutral sentiments towards COVID-19 immunization through vaccines is the highest among others.

Author 1: Haidi Said

Author 2: BenBella S. Tawfik

Author 3: Mohamed A. Makhlouf

Keywords: COVID-19 vaccination; sentiment analysis; Twitter; machine learning; deep learning; natural language processing (NLP)

Paper 59: Validate the Users’ Comfortable Level in the Virtual Reality Walkthrough Environment for Minimizing Motion Sickness

Abstract: Motion sickness is a common scenario for users when they are exposed to a virtual reality (VR) environment. It is due to the conflict that occurs in the brain that tells the user that they are moving in the environment, but the fact is that the user’s body is sitting still causing them to get symptoms of motion sickness like nausea and dizziness. Therefore, motion sickness has become one of the main reasons why users still do not prefer to use VR to enhance their productivity. Motion sickness can be overcome by increasing the user's comfort level of walkthrough in the VR environment. Meanwhile, a popular VR simulation which is widely used in many industries is a walkthrough in a VR environment at a certain speed. This paper is focused on presenting the result of walkthroughs in a VR environment using movement speed and based on frame rates performance and adopting the unified theory of acceptance and use of technology (UTAUT) model construct variables namely performance expectancy (PE) and effort expectancy (EE) to measure the user’s comfort level. A mobile VR, ‘VR Terrain’ application software was developed based on the proposed framework. The application software was tested by 30 users by moving around in a VR environment with 4 different movement speeds that were implemented into four colored gates using a head-mounted display (HMD). A descriptive and coefficient analysis was used to analyze all the data. The blue gate revealed the most comfortable, outperforming all other three gates. Overall, the most suitable speed to use for VR walkthrough is 4.0 km/h. The experiment result may be used to create a parameter for the VR developers to reduce the VR motion sickness effect in the future.

Author 1: Muhammad Danish Affan Anua

Author 2: Ismahafezi Ismail

Author 3: Nur Saadah Mohd Shapri

Author 4: Wan Mohd Amir Fazamin Wan Hamzah

Author 5: Maizan Mat Amin

Author 6: Fazida Karim

Keywords: Virtual reality; motion sickness; head-mounted display; head lean movement; mobile VR; walkthrough technique; UTAUT; frame rate

Paper 60: Fusion Privacy Protection of Graph Neural Network Points of Interest Recommendation

Abstract: For the rapidly developing location-based web recommendation services, traditional point-of-interest(POI) recommendation methods not only fail to utilize user information efficiently, but also face the problem of privacy leakage. Therefore, this paper proposes a privacy-preserving interest point recommendation system that fuses location and user interaction information. The geolocation-based recommendation system uses convolutional neural networks (CNN) to extract the correlation between user and POI interactions and fuse text features, and then combine the location check-in probability to recommend POIs to users. To address the geolocation leakage problem, this paper proposes an algorithm that integrates k-anonymization techniques with homogenized coordinates (KMG) to generalize the real location of users. Finally, this paper integrates location-preserving algorithms and recommendation algorithms to build a privacy-preserving recommendation system. The system is analyzed by information entropy theory and has a high privacy-preserving effect. The experimental results show that the proposed recommendation system has better recommendation performance on the basis of privacy protection compared with other recommendation algorithms.

Author 1: Yong Gan

Author 2: ZhenYu Hu

Keywords: Recommendation algorithms; location protection; graph convolutional neural networks; k-anonymity

Paper 61: Identity Authentication Protocol of Smart Home IoT based on Chebyshev Chaotic Mapping

Abstract: Retracted: After careful and considered review of the content of this paper by a duly constituted expert committee, this paper has been found to be in violation of IJACSA`s Publication Principles. We hereby retract the content of this paper. Reasonable effort should be made to remove all past references to this paper.

Author 1: Jingjing Sun

Author 2: Peng Zhang

Author 3: Xiaohong Kong

Keywords: Chebyshev; chaotic map; internet of things; identity authentication; LAoCCM protocol; key agreement; EDHOC protocol

Paper 62: Hybrid Optimization with Recurrent Neural Network-based Medical Image Processing for Predicting Interstitial Lung Disease

Abstract: One of the dreadful diseases that shortens people's lives is lung disease. There are numerous potentially fatal consequences that can arise from interstitial lung disease, such as: Lung hypertension. This illness doesn't influence your overall blood pressure; instead, it only affects the arteries in your lungs. To prevent mortality, it is essential to accurately diagnose pulmonary illness in patients. Various classifiers, including SVM, RF, MLP, and others, are processed to identify lung disorders. Large datasets cannot be processed by these algorithms, which causes false lung disease identification. A combined new Spider Monkey and Lion algorithm is suggested as a solution to get around these limitations. Images of interstitial lung disease (ILD) were taken for the study from the publicly accessible MedGIFT database. The median filter is employed during the pre-processing step of ILD images to reduce noise and remove undesirable objects. The features are extracted using a hybrid spider Monkey and Lion algorithm. The lungs' damaged and unaffected regions are divided into categories using recurrent neural networks. Several metrics such as accuracy, precision, recall, and f1-score are used to evaluate the performance of the proposed system. The results demonstrate that this technique offers more precision, accuracy, and a higher rate of lung illness detection by processing a large number of computerized tomography representations quickly. When compared to other strategies already in use, the proposed model's accuracy is greater at 99.8%. This method could be beneficial for staging the severity of interstitial lung illness, prognosticating, and forecasting treatment outcomes and survival, determining risk control, and allocation of resources.

Author 1: K. Sundaramoorthy

Author 2: R. Anitha

Author 3: S. Kayalvili

Author 4: Ayat Fawzy Ahmed Ghazala

Author 5: Yousef A.Baker El-Ebiary

Author 6: Sameh Al-Ashmawy

Keywords: Interstitial lung disease; spider monkey lion optimization; recurrent neural network; medical image processing; diagnosis and identification; classification

Paper 63: Study on Tomato Disease Classification based on Leaf Image Recognition based on Deep Learning Technology

Abstract: The utilization of computer vision technology is of the utmost significance in the examination of plant diseases. Research utilizing image processing to investigate plant diseases necessitates the analysis of discernible patterns on plants. Recently, numerous image processing and pattern classification techniques have been employed in the construction of a digital vision system capable of recognizing and categorizing the visual manifestations of plant diseases. Given the abundance of algorithms formulated for the purpose of plant leaf image classification for the detection of plant diseases, it is imperative to assess the accuracy of each algorithm, as well as its potential to identify diverse disease types. The main objective of this study is to explore accurate deep learning architectures that are more effective in deploying and detecting tomato diseases, thus eliminating human error when identifying tomato diseases through visual observation. and get more widespread use. An initial model was constructed from the ground up using a convolutional neural network (CNN), which was trained with 22930 tomato leaf images, and then compared to VGG16, Mobile Net, and Inceptionv3 architectures through a fine-tuning process. The basic CNN model achieved a training accuracy of 90%, whereas the training accuracies of VGG16, Mobile Net, and Inceptionv3 were respectively observed to be 89%, 91%, and 87%. The VGG16 model has a greater computational complexity than other approaches due to its considerable quantity of predefined parameters. Despite to be simpler, MobileNet proved to be the most efficient in terms of accuracy and thus is the most suitable for this research, due to its lightweight structure, fast functioning and adaptability for mobile devices. In contrast to other architectures, the suggested CNN architecture exhibits shallower characteristics, facilitating faster training on the same dataset. This research will provide a solid foundation for future scholars to easily improve the categorization of plant diseases, which is to develop algorithms that are lighter, faster, easier to run, and have higher accuracy.

Author 1: Ji Zheng

Author 2: Minjie Du

Keywords: Deep learning; convolutional neural network; image recognition; plant diseases

Paper 64: Research on Recommendation Model of College English MOOC based on Hybrid Recommendation Algorithm

Abstract: Establishing a reasonable and efficient compulsory education balance index system is very important to boost the all-around of compulsory education development, and then realize the course recommendation for students with different attributes. Based on this, the research aimed at the problems in college English education and evaluation, aimed to establish a college English MOOC education and evaluation system based on the improved neural network recommendation algorithm. The research first constructed the college English MOOC education and evaluation data elements, and then established a genetic algorithm improved neural network algorithm (BP Neural Network Optimization Algorithm Based on Genetic Algorithm, GA-BP), and finally analyzed the effect of the assembled model. These results show that the fitness of the GA-BP model reaches the set expectation when the evolutionary algebra reaches 10 times, and its fitness is 0.6. The corresponding threshold and weight are obtained, and the threshold and weight are substituted into the model. After repeated iterative training, the model finally reached an error of 10-3 when it was trained 12 times, and the expected accuracy was achieved. The R value of each set hovered around 0.97, and the fitting degree was high, which showed that the GA-BP model proposed in the study had a better fitting degree. The difference between the expected value and the output value is mainly distributed in the [-0.08083, 0.06481] interval. To sum up, the GA-BP model proposed in the study has an excellent effect on college English education and evaluation. This evaluation model has a faster learning rate and a higher prediction accuracy and more stable performance.

Author 1: Yifang Ding

Author 2: Jingbo Hao

Keywords: Genetic algorithm; education quality assessment; BP neural network; college English MOOC

Paper 65: Employee Information Security Awareness in the Power Generation Sector of PT ABC

Abstract: Presidential Regulation No. 82 of 2022 demonstrates the Indonesian government's dedication to protecting Vital Information Infrastructure, which has become increasingly susceptible to cyber attacks. Intrusion detections at PT ABC reached 79,575 in 2021, and malware, botnets, targeted attacks, malicious websites/domains, and ransomware attacks may cause considerable financial losses. The implication of these incidents is that employees' awareness of information security is critical, in addition to security technologies like firewalls and monitoring tools. To enhance employees' knowledge of information security, this study aims to evaluate the information security awareness among PT ABC personnel using the HAIS-Q survey instrument alongside ISO/IEC 27001:2013 criteria. The study will provide valuable recommendations to improve the organization's security protocols. This research intends to investigate the correlation between employees' knowledge, attitude, and behavior towards information security. Data was collected through a questionnaire and analyzed using the Pearson Correlation, Cronbach's Alpha, descriptive statistics, linear regression, and Kruskal-Wallis test method. The study findings suggest that the overall information security awareness level among employees is "Good". However, certain areas like internet usage, information handling, asset management, incident reporting, and the use of mobile devices need improvement. To address these areas, the study recommends promoting information security awareness according to employee categories.

Author 1: Ridwan Fadlika

Author 2: Yova Ruldeviyani

Author 3: Zenfrison Tuah Butarbutar

Author 4: Relaci Aprilia Istiqomah

Author 5: Achmad Arzal Fariz

Keywords: Security awareness; data; information; ISO/IEC 27001:2013

Paper 66: ECAH: A New Energy-Aware Coverage Method for Wireless Sensor Networks using Artificial Bee Colony and Harmony Search

Abstract: Wireless Sensor Networks (WSNs) offer diverse applications in the research and commercial fields, such as military applications, medical science, waste management, home automation, habitat monitoring, and environmental observation. WSNs are generally composed of a large number of low-cost, low-power, and multifunctional sensor nodes that sense, process, and communicate data. These nodes are connected by a wireless medium, allowing them to collect and share data with each other. To achieve network coverage in a WSN, a few to thousands of tiny and low-power sensor nodes should be placed in an interconnected manner. Over the last decade, deploying sensor nodes in a WSN to cover a large area has received much attention. Coverage, regarded as an NP-hard problem, is an essential parameter for WSNs that determines how the deployed sensor nodes handle each point of interest. Various algorithms have been proposed to tackle this problem. However, they often come with a trade-off between energy efficiency and coverage rate. Moreover, the scalability of the algorithms needs to be considered for large-scale networks. This paper proposes a novel energy-aware method combining Artificial Bee Colony (ABC) and Harmony Search (HS) algorithms to address the coverage problem in WSN, called ECAH. The proposed ECAH algorithm has been tested with various network scenarios and compared with other existing algorithms. The results show that ECAH outperforms the existing methods in terms of network lifetime, coverage rate, and energy consumption. Additionally, the proposed algorithm is also more robust and efficient as it can adjust to dynamic network environment changes, making it suitable for various network scenarios.

Author 1: ZHOU Bing

Author 2: ZHANG Zhigang

Keywords: Artificial bee colony; coverage; deployment; harmony search; wireless sensor network

Paper 67: Patient Health Monitoring System Development using ESP8266 and Arduino with IoT Platform

Abstract: The Internet of Things (IoT) has emerged as a transformative technology that has revolutionized the field of healthcare. One of the most promising applications of Internet of Things (IoT) in healthcare is patient health monitoring, which allows healthcare providers to remotely monitor patients' health and provide prompt medical attention when needed. This research work focuses on developing an Internet of Things (IoT)-based patient health monitoring system aimed at providing a solution for patients, particularly the elderly, who face the risk of unexpected death due to the lack of medical attention. The proposed system utilizes a heartbeat sensor and an Infrared IR temperature sensor connected to Arduino UNO and Nodemcu, respectively, to monitor the patient's vital signs. The sensors collect the data, which is then sent to an Internet of Things (IoT) web platform via a Wi-Fi connection. The Internet of Things (IoT) platform displays the real-time data of the patient's health status, including the temperature and heartbeat rate, which can be monitored by doctors and nurses. The system is designed to send alerts to healthcare providers in the event of any medical emergency, ensuring that prompt medical attention can be provided to the patient. The significance of this research work lies in its potential to revolutionize the healthcare industry by providing a more efficient and effective means of patient health monitoring. The system can be used to monitor a large number of patients simultaneously, which is particularly beneficial in hospitals with a large patient load. Moreover, it can reduce the workload of healthcare providers, allowing them to focus on other critical tasks. This innovative system has the potential to improve the overall quality of healthcare services and lead to better health outcomes for the society.

Author 1: Jamil Abedalrahim Jamil Alsayaydeh

Author 2: Mohd Faizal bin Yusof

Author 3: Muhammad Zulhakim Bin Abdul Halim

Author 4: Muhammad Noorazlan Shah Zainudin

Author 5: Safarudin Gazali Herawan

Keywords: Patient health monitoring; Internet of Things (IoT); Arduino UNO; Nodemcu ESP8266; thingspeak; wearable device; temperature value; heartbeat value; remotely

Paper 68: Integrated Methodology for Information Security Risk Management using ISO 27005:2018 and NIST SP 800-30 for Insurance Sector

Abstract: The development of Information and Communication Technology (ICT) in the Industrial Revolution 4.0 era shows very fast and disruptive developments that encourage increased use of Information Technology (IT) services within organizations. However, there is a risk of creating vulnerabilities and threats to owned information systems. Plans and strategies are required to implement information security risk management to address vulnerabilities in threat events. This research is a case study of the Enterprise Resource Planning System in the Insurance Sector. The proposed methodologies for integrating information security risk management using ISO/IEC 27005:2018 as a risk management framework and NIST SP 800-30 Rev. 1 as guidance for risk assessments. The risk evaluation stage is the process of comparing the results of the risk analysis with the risk criteria to then determine whether the risk rating is acceptable or tolerable. For risk treatment and control using the ISO/IEC 27002:2022 framework.

Author 1: Arief Prabawa Putra

Author 2: Benfano Soewito

Keywords: Risk management; information security; ISO/IEC 27005; NIST SP 800-30; ISO/IEC 27002

Paper 69: An Effectivity Deep Learning Optimization Model to Traditional Batak Culture Ulos Classification

Abstract: Ulos is one of the Batak culture's traditional heritage fabrics. Ulos cloth is divided into several types, each with a distinct function. Ulos Ragi Hotang, Ulos Pinunsaan, Ulos Tumtuman, Ulos Ragi Hidup, and Ulos Sadum are the five Batak ulos motifs. The Batak ulos motif has evolved over time and is now well-known in other countries. However, many ordinary people have difficulty distinguishing between ulos cloth and other fabrics. This study categorizes the different types of ulos cloth so that it can be used by ordinary people who are unfamiliar with the different types and functions. The Convolutional Neural Network is the method used (CNN). CNN is used to recognize and classify images. CNN's main feature is that it detects feature patches from locations in the input matrix and assembles them into high-level references. The Modular Neural Network (MNN) is then used to break down large and complex computational processes into smaller components, reducing complexity while still producing the desired output. 80% of the data for the training process, 20% for testing. The accuracy value achieved is 97.83%, the loss value is 0.0793, the val loss is 2.1885, and the val accuracy is 0.7429.

Author 1: Rizki Muliono

Author 2: Mayang Septania Iranita

Author 3: Rahmad BY Syah

Keywords: Ulos; classification; convolutional neural network; modular neural network; deep learning

Paper 70: Enhancing Customer Relationship Management Using Fuzzy Association Rules and the Evolutionary Genetic Algorithm

Abstract: The importance of Customer Relationship Management (CRM) has never been higher. Thus, companies are forced to adopt new strategies to focus on customers, given the competitive climate in which they operate. Also, companies have been able to maintain customer data within large databases that contain all information related to customers, thanks to the tremendous technological development seen recently. Multilevel quantitative association mining is a significant field for achieving motivational associations between data components with multiple abstraction levels. This paper develops a methodology to support CRM to improve the relationship between retail companies and their customers in the retail sector to retain existing customers and attract more new customers, by applying data mining techniques using the genetic algorithm through which an integrated search is performed. The proposed model can be implemented because the proposed model does not need the minimum levels of support and trust required by the user, and it has been confirmed that the algorithm proposed in this research can powerfully create non-redundant fuzzy multi-level association rules, according to the results of these experiments.

Author 1: Ahmed Abu-Al Dahab

Author 2: Riham M Haggag

Author 3: Samir Abu-Al Fotouh

Keywords: Customer relationship management (CRM); fuzzy association rule mining; multilevel association rule; quantitative data mining

Paper 71: Leader-follower Optimal Control Method for Vehicle Platoons to Improve Fuel Eﬃciency

Abstract: The automotive industry has experienced swift development with the rapid growth of the economy. The increasing number of vehicles has led to deteriorating road traffic conditions, increased consumption of nonrenewable energy, and excessive vehicle emissions. To tackle the problems of fuel efficiency and safety control for vehicle platoons, this study suggests a novel leader-follower optimal control method for vehicle platoons to improve fuel efficiency. Depending on where vehicles are in a platoon, they are classified into two categories: the leader vehicle and follower vehicles. The differing driving circumstances of these two types of vehicles are considered in this paper by using various control methods. On the one hand, the leader vehicle uses an approximate fuel consumption model to improve computational efficiency. At the same time, speed and acceleration are constrained to obtain the speed optimization curve. The follower vehicle uses a distributed receding horizon control method, which calculates the vehicle's speed optimization profile online. On the other hand, a linear following model is used to prevent collisions between vehicles and ensure the safety of the vehicle platoon. The simulation experiment has demonstrated that this speed optimization control method can reduce the fuel consumption of the vehicle platoon and ensure the safety of the vehicles.

Author 1: Zhigang Li

Author 2: Yushi Guo

Author 3: Hua Wang

Author 4: Jianyong Li

Author 5: Yuye Xie

Author 6: Jingyu Liu

Keywords: Vehicle platoon speed optimization control; leader vehicles; follower vehicles; distributed re-ceding horizon control method; vehicle platoon fuel consumption control

Paper 72: A Review of the Recent Progress on Crowd Anomaly Detection

Abstract: Surveillance videos are crucial in imparting public security, reducing or avoiding the accidents that occur from anomalies. Crowd anomaly detection is a rapidly growing research field that aims to identify abnormal or suspicious behavior in crowds. This paper provides a comprehensive review of the state-of-the-art in crowd anomaly detection and, different taxonomies, publicly available datasets, challenges, and future research directions. The paper first provides an overview of the field and the importance of crowd anomaly detection in various applications such as public safety, transportation, and surveillance. Secondly, it presents the components of crowd anomaly detection and its different taxonomies based on the availability of labels, and the type of anomalies. Thirdly, it presents the review of the recent progress of crowd anomaly detection. The review also covers publicly available datasets commonly used for evaluating crowd anomaly detection methods. The challenges faced by the field, such as handling variability in crowd behavior, dealing with large and complex data sets, and addressing the imbalance of data, are discussed. Finally, the paper concludes with a discussion of future research directions in crowd anomaly detection, including integrating multiple modalities, addressing privacy concerns, and addressing crowd monitoring systems’ ethical and legal implications.

Author 1: Sarah Altowairqi

Author 2: Suhuai Luo

Author 3: Peter Greer

Keywords: Crowd anomaly detection; advanced computer science; intelligent systems; video surveillance application; machine learning

Paper 73: Improving Brain Tumor Segmentation in MRI Images through Enhanced Convolutional Neural Networks

Abstract: Achieving precise tumor segmentation is essential for accurate diagnosis. Since brain tumors segmentation require a significant training process, reducing the training time is critical for timely treatment. The research focuses on enhancing brain tumor segmentation in MRI images by using Convolutional Neural Networks and reducing training time by using MATLAB's GoogLeNet, anisotropic diffusion filtering, morphological operation, and sector vector machine for MRI images. The proposed method will allow for efficient analysis and management of enormous amounts of MRI image data, the earliest practicable early diagnosis, and assistance in the classification of normal, benign, or malignant patient cases. The SVM Classifier is used to find a cluster of tumors development in an MR slice, identify tumor cells, and assess the size of the tumor that appears to be present in order to diagnose brain tumors. The proposed method is evaluated using a dataset from Figshare that includes coronal, sagittal, and axial views of images taken with a T1-CE MRI modality. The accuracy of 2D tumor detection and segmentation are increased, enabling more 3D detection, and achieving a mean classification accuracy of 98% across system records. Finally, a hybrid approach of GoogLeNet deep learning algorithm and Convolution Neural Network- Support Vector Machines (CNN-SVM) deep learning is performed to increase the accuracy of tumor classification. The evaluations show that the proposed technique is significantly more effective than those currently in use. In the future, enhancement of the segmentation using artificial neural networks will help in the earlier and more precise detection of brain tumors. Early detection of brain tumors can benefit patients, healthcare providers, and the healthcare system as a whole. It can reduce healthcare costs associated with treating advanced stage tumors, and enables researchers to better understand the disease and develop more effective treatments.

Author 1: Kabirat Sulaiman Ayomide

Author 2: Teh Noranis Mohd Aris

Author 3: Maslina Zolkepli

Keywords: MRI brain tumor; anisotropic; segmentation; SVM classifier; convolutional neural network

Paper 74: Rain Streaks Removal in Images using Extended Generative Adversarial-based Deraining Framework

Abstract: The visual quality of photographs and videos can be negatively impacted by various weather conditions, such as snow, haze, or rain, affecting the quality of the images and videos. Such impacts may greatly affect outdoor vision systems that rely on image/video data. It has recently drawn a lot of interest to remove rain streaks from a single image. Several deep learning-based methods have been introduced to address the issue of removing rain streaks from a single image. Still, the efficiency of rain streak removal with enhanced quality is challenging. Hence, a novel deep-learning method is introduced for rain streak removal. The proposed Extended Generative Adversarial based De-raining (Ex_GADerain) is the enhanced version of a traditional Generative adversarial network (GAN). The proposed Ex_GADerain introduced a Self-Attention based Convolutional Capsule Bidirectional Network (SA-CCapBiNet) based generator for enhancing the rain streaks removal process. Also, the loss function estimation using the adversarial loss and the mean absolute error loss minimizes the information loss during training. The minimal information loss enhances the generalization capability of Ex_GADerain, and hence the enhanced performance is acquired. The quality assessment of a derained image based on various assessment measures like SSIM, PSNR, RMSE, and DSSIM improved performance compared to the conventional rain streak removal methods. The maximal SSIM and PSNR acquired by the Ex_GADerain are 0.9923 and 26.7052, respectively. The minimal RMSE and DSSIM acquired by the Ex_GADerain are 0.9367 and 0.0051, respectively.

Author 1: Subbarao Gogulamudi

Author 2: V. Mahalakshmi

Author 3: Indraneel Sreeram

Keywords: Deep learning; rain streaks removal; image generation; quality measure; capsule network; adversarial learning

Paper 75: A PSL-based Approach to Human Activity Recognition in Smart Home Environments

Abstract: Human activity recognition is widely used in smart cities, public safety and other fields, especially in smart home systems where it has a pivotal role. The study addresses the shortcomings of Markov logic networks for human activity recognition and proposes a human activity recognition method in smart home scenarios - an activity recognition framework based on Probabilistic Soft Logic (PSL). The framework is able to deal with logical uncertainty problems and provides expression and inference mechanisms for data uncertainty problems on this basis. The framework utilizes Deng entropy evidence theory to provide an evaluation method for sensor event uncertainty, and combines event calculus for activity modeling. Comparing the PSL method with three other common recognition methods, Ontology, Hidden Markov Model (HMM), and Markov logic network, on a public dataset, it was found that the PSL method has a much better ability to handle data uncertainty than the other three algorithms. The average recognition rates on the ADL and ADL-E sub datasets were 82.87% and 80.33%, respectively. In experiments to verify the ability of PSL to handle temporal complexity, PSL showed the least significant decrease in the average recognition rate and maintained an average recognition rate of 81.02% in the presence of concurrent and alternating activities. The human activity recognition method based on PSL has a better performance in handling both data uncertainty and temporal complexity.

Author 1: Yan Li

Keywords: Human activity recognition; probabilistic soft logic; MAP inference; temporal complexity; data uncertainty

Paper 76: Design of Virtual Experiment Teaching of Inorganic Chemistry in Colleges and Universities Based on Unity3D

Abstract: Focusing on the high-cost issues and many risk and uncontrollable factors in chemical experiment teaching in modern colleges and universities, a virtual reality inorganic chemistry (IC) simulation experiment strategy based on Unity3D technology is studied. Starting from the needs of IC experiment teaching, a collision detection algorithm (CDA) between the experimental environment and the virtual scene with Unity3D as the main technical framework is designed. The results show that the collision detection (CD) time of the CDA designed in the research is 99ms, the detection average value is 95.29%, and the accuracy variance is 0.021. The above values are better than the same type of algorithm. This shows that the CD accuracy and efficiency are higher, and the performance is stronger. In addition, the virtual chemistry experiment designed in the research can significantly improve the students’ using attitude from the three main aspects of perceived ease of use (PEU), immersion and interactivity, and then enhance the students’ learning effect. Therefore, the method of virtual experiment teaching of IC in colleges and universities based on Unity3D is effective and provides a new idea for modern teaching reform.

Author 1: Xia Hu

Keywords: Unity3D; inorganic chemistry; virtual reality; collision detection algorithm

Paper 77: The Effective 3D MRI Reconstruction Method Driven by the Fusion Strategy in NSST Domain

Abstract: The 3D reconstruction of medical images plays an important role in modern clinical diagnosis. Although the analytic-based, the iterative-based and the deep learning-based methods have been popularly used, there are still many problems to deal with. The analysis-based methods are not accurate enough, the iteration-based methods are computationally intensive, and the deep learning based methods are heavily dependent on the training of the data. To solve the default that only the single scan sequence is included in the traditional methods, a reconstruction method driven by the non-subsampled shearlet transform (NSST) and the algebraic reconstruction technique (ART) is proposed. Firstly, the multiple magnetic resonance imaging (MRI) sequences are decomposed into high-frequency and low-frequency components by NSST. Secondly, the low-frequency parts are fused with the weighted average fusion scheme and the high-frequency parts are fused with the weighted coefficient scheme that guided by the regional average gradient and energy. Finally, the 3D reconstruction is performed by using the ART algorithm. Compared with the traditional reconstruction methods, the proposed method is able to capture more information from the multiple MRI sequences, which makes the reconstruction results much clearer and more accurate. By comparing with the single-sequence reconstruction model without fusion, the experiments fully prove the accuracy and effectiveness.

Author 1: Jin Huang

Author 2: Lei Wang

Author 3: Muhammad Tahir

Author 4: Tianqi Cheng

Author 5: Xinping Guo

Author 6: Yuwei Wang

Author 7: ChunXiang Liu

Keywords: Multiple magnetic resonance imaging; 3D reconstruction; non-subsampled shearlet transform; the algebraic reconstruction

Paper 78: Integrating Dropout Regularization Technique at Different Layers to Improve the Performance of Neural Networks

Abstract: In many facial expression recognition models it is necessary to prevent overfitting to check no units (neurons) depend on each other. Therefore, dropout regularization can be applied to ignore few nodes randomly while processing the remaining neurons. Hence, dropout helps dealing with overfitting and predicts the desired results with more accuracy at different layers of the neural network like ‘visible’, ‘hidden’ and ‘convolutional’ layers. In neural networks there are layers like dense, fully connected, convolutional and recurrent (LSTM- long short term memory). It is possible to embed the dropout layer with any of these layers. Model drops the units randomly from the neural network, meaning model removes its connection from other units. Many researchers found dropout regularization a most powerful technique in machine learning and deep learning. Dropping few units (neurons) randomly and processing the remaining units can be considered in two phases like forward and backward pass (stages). Once the model drops few units randomly and select ‘n’ from the remaining units it is obvious that weight of the units could change during processing. It must be noted that updated weight doesn’t reflect on the dropped units. Dropping and stepping-in few units seem to be very good process as those units which step-in will represent the network. It is assumed to have maximum chance for the stepped-in units to have less dependency and model gives better results with higher accuracy.

Author 1: B. H. Pansambal

Author 2: A. B. Nandgaokar

Keywords: Convolutional layer; visible layer; hidden layer; dropout regularization; long short term memory (LSTM); deep learning; facial expression recognition

Paper 79: Comparison Review on Brain Tumor Classification and Segmentation using Convolutional Neural Network (CNN) and Capsule Network

Abstract: Malignant brain glioma is considered as one of the deadliest cancer diseases that has a higher fatality rate than the survival rate. In terms of brain glioma imaging and diagnosis, the processes of detection and segmentation are manually done by the experts. However, with the advancement of artificial intelligence, the implementation of these tasks using deep learning provides an efficient solution to the management of brain glioma diagnosis and patient treatment. Deep learning networks are responsible for detecting, segmenting, and interpreting the tumors with high accuracy and repeatability so that the appropriate treatment planning can be offered to the patient. This paper presents a comparison review between two state of the art deep learning networks namely convolutional neural network and capsule network in performing brain glioma classification and segmentation tasks. The performance of each published method is discussed along with their advantages and disadvantages. Next, the related constraints in both networks are outlined and highlighted for future research.

Author 1: Nurul Fatihah Binti Ali

Author 2: Siti Salasiah Mokri

Author 3: Syahirah Abd Halim

Author 4: Noraishikin Zulkarnain

Author 5: Ashrani Aizuddin Abd Rahni

Author 6: Seri Mastura Mustaza

Keywords: Deep learning; convolution neural network (CNN); capsule network; segmentation; classification; brain glioma

Paper 80: Fuzzy Reasoning based Reliability Fault Prediction of CNC Machine Tools

Abstract: CNC machine tools are the infrastructure of the manufacturing industry, and many fields cannot do without them. This paper studies the fault data of a series of CNC machine tools, and predicts the fault level based on the activity parameters of Gutenberg Richter curve and fuzzy information theory. Apply the Gutenberg Richter curve model to the reliability analysis of CNC machine tools, and use this model to fit the curves separately. Fit the activity parameters of each stage with curves, and the results show that the b value can reflect the fault activity frequency of CNC machine tools. Due to the correlation and fuzziness between system faults, it is more appropriate to use a fuzzy neural network with strong adaptability and good learning ability, which can easily adjust parameters, and can express a more complex, high-dimensional nonlinear system through fewer conditions. The use of fuzzy reasoning can link the nonlinear relationship between fault level, b-value, and N-value. Analyze the error between the predicted fault level and the original level, and the small error indicates that the model has good predictive ability. Applying this predictive ability to the reliability research of CNC machine tools will yield good results.

Author 1: Jie Yu

Author 2: Tiebin Wang

Author 3: Weidong Wang

Author 4: Gege Zhao

Author 5: Yue Yao

Keywords: CNC machine tools; reliability; fuzzy inference; fault prediction

Paper 81: Using Machine Learning Algorithm as a Method for Improving Stroke Prediction

Abstract: Having sudden strokes has had a very negative impact on all aspects in society to the point that it attracted efforts for better improvement and management of stroke diagnosis. Technological advancement also had an impact on the medical field such that nowadays caregivers have better options for taking care of their patients by mining and archiving their medical records for ease of retrieval. Furthermore, it is quite essential to understand the risk factors that make a patient more susceptible to strokes, thus there are some factors that make stroke prediction much easier. This research offers an analysis of the factors that enhance the stroke prediction process based on electronic health records. The most important factors for stroke prediction will be identified using statistical methods and Principal Component Analysis (PCA). It has been found that the most critical factors affecting stroke prediction are the age, average glucose level, heart disease, and hypertension. A balanced dataset is used for the model evaluation which was created by sub-sampling since the dataset for stroke occurrence is already highly imbalanced. In this study, seven different machine learning algorithms are implemented: Naïve Bayes, SVM, Random Forest, KNN, Decision Tree, Stacking, and majority voting to train on the Kaggle dataset to predict occurrence of stroke in patients. After preprocessing and splitting the dataset into training and testing sub-datasets, these proposed algorithms were evaluated according to accuracy, f1 score, recall value, and precision value. The NB classifier achieved the lowest accuracy level (86%), whereas the rest of the algorithms achieved similar accuracies 96%, f1 scores 0.98, precision 0.97, and recall 1.

Author 1: Nojood Alageel

Author 2: Rahaf Alharbi

Author 3: Rehab Alharbi

Author 4: Maryam Alsayil

Author 5: Lubna A. Alharbi

Keywords: Stroke prediction; machine learning; PCA; decision tree; KNN; majority voting; Naïve Bayes

Paper 82: Deep Learning Localization Algorithm Integrating Attention Mechanism in Database Information Query

Abstract: This study aims to solve the problems of traditional indoor car search positioning technology in terms of positioning accuracy and functionality. Based on database technology and deep learning technology, an LSTM model with attention mechanism was established. This model can simultaneously extract temporal and spatial features, and use attention mechanism for feature importance recognition. The entire positioning model has been designed as a triple functional entrance that includes positioning, car storage, and reverse car search, enhancing the user's coherent experience. The data results show that the root mean square error of the LSTM (Attention) model designed in the study is 0.216, and the variance is 0.092. Among similar positioning models, the index value is the smallest, while the CDF line rises the fastest and the maximum value is the highest. The research conclusion indicates that the LSTM (Attention) indoor positioning model designed in this study has better computational performance and can help users achieve more accurate positioning and vehicle navigation.

Author 1: Yang Li

Author 2: Xianghui Hui

Author 3: Xiaolei Wang

Author 4: Fei Yin

Keywords: LSTM; attention mechanism; positioning; database

Paper 83: Develop an Olive-based Grading Algorithm using Image Processing

Abstract: Olives come in a number of external and internal varieties. The Shengeh kind, which is available in three colours—green, brown, and black—was chosen at random by the researchers to ensure that the sample was diverse. To avoid discoloration throughout the experiment, 150 healthy olives were harvested and stored correctly. These olives had not been subjected to any external harm, such as crushing or milling. The particular kind of olives were kept chilled at 2°C and preserved in water. This study investigates the possibility of grading Shengeh cultivars from olives that have different uses, based on color using image processing. After preparing images of olives using MATLAB software and image processing techniques, olives are graded based on their color in three categories: immature with green, semi-ripe with brown, and ripe with black. The results showed that image processing technology can be used to grade olives of the Shengeh type in terms of their ripeness as a single-color grain with acceptable accuracy. The HSV color space is one of the best color spaces to separate the colors of the olive cultivar. The accuracy of the software for detecting olives with the mentioned degrees is 98%, 96%, and 100%, respectively.

Author 1: Dongliang Jin

Keywords: Image processing; grading; color; olive; MATLAB; HSV

Paper 84: Intelligent Abnormal Residents’ Behavior Detection in Smart Homes for Risk Management using Fuzzy Logic Algorithm

Abstract: In recent years, the population of sick and elderly people who are alone and need care has increased. This issue increases the need to have a smart home to be aware of the patient's condition. Identifying the patient's activity using sensors embedded in the environment is the first step to reach a smart home where the people around the patient can leave the patient alone at home with less worry. In literature, a variety of methods for detecting the performance of users in the smart home are discussed. In this study, a method for abnormal behavior detection and identifying the level of risk is proposed, in which fuzzy logic is used in cases such as when the activity start. Experimental results demonstrates that the proposed method achieved satisfied performance with 90% accuracy rate that presented better results compared to other existing methods.

Author 1: Bo Feng

Author 2: Lili Miao

Author 3: HuiXiang Liu

Keywords: Smart home; abnormal detection; behavior analysis; activity recognition; elderly people; fuzzy logic

Paper 85: BREPubSub: A Secure Publish-Subscribe Model using Blockchain and Re-encryption for IoT Data Sharing Management

Abstract: As a result of the incredible growth and diversity of IoT systems and applications over the past several years, an enor-mous amount of sensing data has been generated, which is critical for developing IoT-based intelligent systems. So far, it has taken a significant amount of time and money to collect sufficient sensing data for these smart systems leading to demands of sharing or exchanging available and valuable data to reduce the time and money spent on the data collection process. However, ensuring the data sharing process’s integrity, security, and fairness is fraught with challenges. This paper proposes a Blockchain-based model that supports a secure publish-subscribe protocol for data sharing management by addressing three criteria such as confidentiality, integrity, and availability. In addition, the proposed model adopts a re-encryption technique to optimize shared data storage with multiple users and enhance the security of the data exchange process in a transparent and public environment like Blockchain. We have developed a DApp to demonstrate the feasibility of our design and evaluate its performance.

Author 1: Hoang-Anh Pham

Keywords: Publish-subscribe; blockchain; re-encryption; IoT data sharing

Paper 86: A Review of Trending Crowdsourcing Topics in Software Engineering Highlighting Mobile Crowdsourcing and AI Utilization

Abstract: Today’s modern technologies and requirements make the utilization of crowdsourcing more viable and applicable. It is one of the problem-solving models that can be used in various domains to reduce costs and time. It is also an excellent way to find new and different ideas and solutions. This paper studies the use of crowdsourcing in software engineering and reveals adequate details to highlight its significance. A few recent literature reviews have been published to address specific topics or study general attributes of papers in crowdsourced software engineering. This paper, however, explores all recent publications related to software and crowdsourcing to find the trends and highlight mobile and AI usage in software crowdsourcing. The findings of this paper show that most research papers are in the areas of software management and software verification and validation. The results also reveal that machine learning and data mining techniques are predominant in software management crowdsourcing and software verification and validation. Furthermore, this study shows that the methods and techniques used in general crowdsourcing apply to mobile crowdsourcing except in mobile testing, where there is a need for clustering and prioritization of test reports.

Author 1: Mohammed Alghasham

Author 2: Mousa Alzakan

Author 3: Mohammed Al-Hagery

Keywords: Software engineering; crowdsourcing; mobile crowdsourcing; software management; software verification and validation

Paper 87: Evaluation of Wood Species Identification Using CNN-Based Networks at Different Magnification Levels

Abstract: Wood species identification (WoodID) is a crucial task in many industries, including forestry, construction, and furniture manufacturing. However, this process currently requires highly trained individuals and is time-consuming. With the recent advances in machine learning and computer vision techniques, automatic WoodID using macro-images of cross-section wood has gained attention. Nevertheless, existing works have been evaluated on ad-hoc datasets with pre-fixed magnification levels. To address this issue, this paper proposes an evaluation of deep learning-based methods for WoodID on multiple datasets with varying magnification levels. Several popular Convolutional Neural Networks, including DenseNet, ResNet50, and MobileNet, were examined to identify the best network and magnification levels. The experiments were conducted on five datasets with different magnifications, including a self-collected dataset and four existing ones. The results demonstrate that the DenseNet121 network achieved superior accuracy and F1-Score on the 20X dataset. The findings of this study provide useful insights into the development of automatic WoodID systems for practical applications.

Author 1: Khanh Nguyen-Trong

Keywords: Wood species identification; convolutional neural network; ResNet50; DensNet

Paper 88: A Review of Milgram and Kishino’s Reality-Virtuality Continuum and a Mathematical Formalization for Combining Multiple Reality-Virtuality Continua

Abstract: We explore in this paper theoretical contributions that are related to Milgram and Kishino’s Reality Virtuality Continuum by conducting a systematic literature review. From this study, we draw inspiration for our proposed mathematical formalization of combining multiple Reality-Virtuality Continua in a single, mixed reality experience. Also, we provide a definition for XR transition protocol. To complete our contribution, we discuss two potential examples that will exemplify our formalization and identify future work to be addressed.

Author 1: Cristian Pamparau

Keywords: Systematic literature review; reality-virtuality continuum; mixed reality; transitional interfaces; mathematical for-malization; XR transition protocol

Paper 89: Evolutionary Design of a PSO-Tuned Multigene Symbolic Regression Genetic Programming Model for River Flow Forecasting

Abstract: The earth’s population is growing at a rapid rate, while the availability of water resources remains limited. Water is required for various purposes, including drinking, agriculture, industry, recreation, and development. Accurate forecasting of river flows can have a significant economic impact, particularly in agricultural water management and planning during water resource scarcity. Developing precise river flow forecasting models can greatly improve the management of water resources in many countries. In this study, we propose a two-phase model for predicting the flow of the Blackwater river located in the South Central United States. In the first phase, we use Multigene Symbolic Regression Genetic Programming (MG-GP) to develop a mathematical model. In the second phase, Particle Swarm Optimization (PSO) is employed to fine-tune the model parameters. Fine-tuning the MG-GP parameters improves the prediction accuracy of the model. The newly fine-tuned model exhibits 96% and 94% accuracy in training and testing cases, respectively.

Author 1: Alaa Sheta

Author 2: Amal Abdel-Raouf

Author 3: Khalid M. Fraihat

Author 4: Abdelkarim Baareh

Keywords: River flow; forecasting; genetic programming; evolutionary computation; particle swarm optimization

Paper 90: Autonomous Motion Planning for a Differential Robot using Particle Swarm Optimization

Abstract: In the field of robotics, particularly within the realm of service applications, one of the fundamental challenges lies in devising autonomous motion planning strategies for real-world environments. Addressing this issue necessitates the management of numerous variables, with the primary goal of enabling the robot to circumnavigate obstacles, attain its target destination in the most efficient manner, and adhere to the shortest possible route while prioritizing safety. Furthermore, the robot’s control mechanisms must exhibit stability, precision, and swift respon-siveness. Prompted by these requirements, this paper explores the utilization of Particle Swarm Optimization (PSO) in conjunction with a Proportional-Integral-Derivative (PID) controller to devise a motion planning strategy for a differential robot operating in a multifaceted real-world setting. The proposed control system is implemented using an ESP32 microcontroller, which serves as the foundation for the robot’s motion planning and execution capabilities. Through a series of simulations, the efficacy of the suggested approach is demonstrated, emphasizing its potential as a robust solution for addressing the complex challenge of autonomous motion planning in real-world environments.

Author 1: Fredy Martinez

Author 2: Angelica Rendon

Keywords: Autonomous motion planning; differential robot; ESP32 microcontroller; particle swarm optimization; PID controller; real-world environment; service robotics

Paper 91: Software Effort Estimation using Machine Learning Technique

Abstract: Software engineering effort estimation plays a significant role in managing project cost, quality, and time and creating software. Researchers have been paying close attention to software estimation during the past few decades, and a great amount of work has been done utilizing a variety of machine-learning techniques and algorithms. In order to better effectively evaluate predictions, this study recommends various machine learning algorithms for estimating, including k-nearest neighbor regression, support vector regression, and decision trees. These methods are now used by the software development industry for software estimating with the goal of overcoming the limitations of parametric and conventional estimation techniques and advancing projects. Our dataset, which was created by a software company called Edusoft Consulted LTD, was used to assess the effectiveness of the established method. The three commonly used performance evaluation measures, mean absolute error (MAE), mean squared error (MSE), and R square error, represent the base for these. Comparative experimental results demonstrate that decision trees perform better at predicting effort than other techniques.

Author 1: Mizanur Rahman

Author 2: Partha Protim Roy

Author 3: Mohammad Ali

Author 4: Teresa Gonc¸alves

Author 5: Hasan Sarwar

Keywords: Software effort estimation; K-nearest neighbor re-gression; machine learning; decision tree; support vector regression

Paper 92: An Approach to Hyperparameter Tuning in Transfer Learning for Driver Drowsiness Detection Based on Bayesian Optimization and Random Search

Abstract: Driver drowsiness is a critical factor in road safety, and developing accurate models for detecting it is essential. Transfer learning has been shown to be an effective technique for driver drowsiness detection, as it enables models to leverage large, pre-existing datasets. However, the optimization of hyper-parameters in transfer learning models can be challenging, as it involves a large search space. The core purpose of this research is on presenting an approach to hyperparameter tuning in transfer learning for driving fatigue detection based on Bayesian optimization and Random search algorithms. We examine the efficiency of our approach on a publicly available dataset using transfer learning models with the MobileNetV2, Xception, and VGG19 architectures. We explore the impact of hyperparameters such as dropout rate, activation function, the number of units (the number of dense nodes), optimizer, and learning rate on the transfer learning models’ overall performance. Our experiments show that our approach improves the performance of the transfer learning models, obtaining cutting-edge results on the dataset for all three architectures. We also compare the efficiency of Bayesian optimization and Random search algorithms in terms of their ability to find optimal hyperparameters and indicate that Bayesian optimization is more efficient in finding optimal hyperparameters than Random search. The results of our study provide insights into the importance of hyperparameter tuning for transfer learning-based driver drowsiness detection using different transfer learning models and can guide the selection of hyperparameters and models for future studies in this field. Our proposed approach can be applied to other transfer learning tasks, making it a valuable contribution to the field of ML.

Author 1: Hoang-Tu Vo

Author 2: Hoang Tran Ngoc

Author 3: Luyl-Da Quach

Keywords: Hyperparameter tuning; driver drowsiness detection; transfer learning; Bayesian optimization; Random search

Paper 93: Discovering COVID-19 Death Patterns from Deceased Patients: A Case Study in Saudi Arabia

Abstract: COVID-19 is a serious infection that cause severe injuries and deaths worldwide. The COVID-19 virus can infect people of all ages, especially the elderly. Furthermore, elderly who have co-morbid conditions (e.g., chronic conditions) are at an increased risk of death. At the present time, no approach exists that can facilitate the characterization of patterns of COVID-19 death. In this study, an approach to identify patterns of COVID- 19 death efficiently and systematically is applied by adapting the Apriori algorithm. Validation and evaluation of the proposed approach are based on a robust and reliable dataset collected from Health Affairs in the Makkah region of Saudi Arabia. The study results show that there are strong associations between hypertension, diabetes, cardiovascular disease, and kidney disease and death among COVID-19 deceased patients.

Author 1: Abdulrahman Alomary

Author 2: Tarik Alafif

Author 3: Abdulmohsen Almalawi

Author 4: Anas Hadi

Author 5: Faris Alkhilaiwi

Author 6: Yasser Alatawi

Keywords: COVID-19; association rules; Apriori algorithm; patterns; death; chronic diseases

Paper 94: Highly Accurate Deep Learning Model for Olive Leaf Disease Classification: A Study in Tacna-Per´u

Abstract: Deep learning applied to computer vision has different applications in agriculture, medicine, marketing, meteorology, etc. In agriculture, plant diseases can cause significant yield and quality losses. The treatment of these diseases depends on accurate and rapid classification. Olive leaf diseases are a problem that threatens the crop quality of olive growers. The objective of this work was to classify olive leaf diseases with Deep Learning in olive crops of the La Yarada-Los Palos area in the Tacna region, Peru. Disease classification is a critical task, nevertheless, for the most common diseases in the region: virosis, fumagina, and nutritional deficiencies, there is no dataset to train deep learning models. Due to the latter, a novel dataset of RGB olive leaf images is elaborated and published. Then, an extensive comparative ex-perimental study was conducted using all possible configurations of Learning from Scratch, Transfer Learning, Fine-Tuning, and Data Augmentation state-of-the-art methods to train a modified VGG16 architecture for the classification of Olive Leaf Diseases. It was demonstrated experimentally: (i) The ineffectiveness of Data Augmentation when the model Learning from Scratch, (ii) A high improvement by using Transfer Learning vs Learning from Scratch, (iii) Similar performance using Transfer Learning vs Transfer Learning + Fine-Tuning vs Transfer Learning + Data Augmentation, and (iv) Very high improvement using Transfer Learning + Fine-Tuning + Data Augmentation. This led us to a Deep Learning Model with an accuracy of 100%, 99.93%, and 100% in the training, validation, and test sets and F1-Score on the validation set of 1, 0.9901, and 0.9899 in the Nutritional Deficiences, Fumagina, and Virosis olive leaf diseases respectively. Replication of the results is ensured by publishing the novel dataset and the final model on GitHub.

Author 1: Erbert F. Osco-Mamani

Author 2: Israel N. Chaparro-Cruz

Keywords: Olive; leaf diseases; disease classification; deep learning; data augmentation; transfer learning; fine-tuning; VGG16

Paper 95: Enhanced MQTT Architecture for Smart Supply Chain

Abstract: In industry 4.0, the use of smart supply chains has become necessary in order to overcome the shortcomings of traditional supply chains, such as overstocking, delivery delays, and stock out. However, the use of smart supply chains has introduced new security challenges because of the internet of things (IOT) constraint nature. Thus, the problem raised is ensuring the supply chain security requirements while taking into consideration the properties of the constraint environment. For this purpose, this paper aims to strengthen the authentication and data transmitting processes of the Message Queuing Telemetry Transport(MQTT) protocol, as the most used protocol for communication in the IOT environment, using blockchain and smart contracts. The new MQTT architecture allows to avoid a single point of failure, to ensure data immutability and to automatize the authentication mechanism as well as the publishing and the subscribing processes. In addition, the use of a one-time password (OTP) instead of a permanent one is another security measure used to protect the architecture from identity spoofing. The new architecture comprises three phases: Registration, Connection, and Publishing. Each phase is automatically controlled by a smart contract. For attack simulation tests, the smart contracts are implemented in a remix environment. The results of the simulation tests show that the new architecture is robust and resistant to different attacks.

Author 1: Raouya AKNIN

Author 2: Youssef Bentaleb

Keywords: Smart supply chain; internet of things; MQTT protocol; blockchain; smart contracts; Ethereum; solidity; one-time password

Paper 96: Hybrid Machine Learning-Based Approach for Anomaly Detection using Apache Spark

Abstract: Over the past few decades, the volume of data has increased significantly in both scientific institutions and universities, with a large number of students enrolled and a high volume of related data. Furthermore, network traffic has increased with post-pandemic and the use of online learning. Therefore, processing network traffic data is a complex and challenging task that increases the possibility of intrusions and anomalies. Traditional security systems cannot deal with such high-speed and big data traffic. Real-time anomaly detection should be able to process data as quickly as possible to detect abnormal and malicious data. This paper proposes a hybrid approach consisting of supervised and unsupervised learning for anomaly detection based on the big data engine Apache Spark. Initially, the k-means algorithm was implemented in Sparks MLlib for clustering network traffic, then for each cluster, K-nearest neighbors algorithm (KNN) was implemented for classification and anomaly detection. The proposed model was trained and validated against a real dataset from Ibn Zohr University. The results indicate that the proposed model outperformed other well-known algorithms in detecting anomalies based on the aforementioned dataset. The experimental results show that the proposed hybrid approach can reach up to 99.94 % accuracy using the k-fold cross-validation method in the complete dataset with all 48 features.

Author 1: Hanane Chliah

Author 2: Amal Battou

Author 3: Maryem Ait el hadj

Author 4: Adil Laoufi

Keywords: Anomaly detection; big data; Apache Spark; k-means; KNN

Paper 97: Egypt Monuments Dataset version 1: A Scalable Benchmark for Image Classification and Monument Recognition

Abstract: The success of machine learning (ML) as well as deep learning (DL) depends largely on data availability and quality. The system’s performance is frequently more affected by the amount and quality of its training data than by its architecture and training specifics. Consequently, demand exists for challenging datasets that both precisely measure performance and present unique challenges with real-world applications. The Egypt Monuments Dataset v1 (EGYPT-v1) is introduced as a new scalable benchmark for fine-image classification (IC) and object recognition (OR) in the domain of ancient Egyptian monuments. EGYPT-v1 dataset is by far the world’s first large specified such dataset to date, with over seven thousand images and 40 distinct instance labels. The dataset composes different categories of monuments such as pyramids, temples, mummies, statues, head statues, bust statues, heritage sites, palaces and shrines. Several advanced deep network architectures were tested to appraise the classification difficulty in the EGYPT-v1 dataset, namely ResNet50, Inception V3, and LeNet5 models. The models achieved accuracy rates as follows: 99.13%, 90.90%, and 92.64%, respectively. The dataset was predominantly created by manually collecting images from the popular global online video-sharing and social media platform, Youtube, as well as WATCHiT, Egypt’s top streaming entertainment service. Additionally, Wikimedia Commons, the largest crowdsourced media repository in the world, was used as a secondary source of images. The images that comprise the dataset can be accessed on the GitHub repository https://github.com/mennatallahhassan/egypt-monuments-dataset.

Author 1: Mennat Allah Hassan

Author 2: Alaa Hamdy

Author 3: Mona Nasr

Keywords: Deep learning; landmark datasets; landmark recognition; monument datasets; monument recognition

Paper 98: Exploring the Joint Potential of Blockchain and AI for Securing Internet of Things

Abstract: The emergence of the Internet of Things (IoT) has revolutionized the way we interact with the physical world. The rapid growth of IoT devices has led to a pressing need for robust security measures. Two promising approaches that can enhance IoT security are blockchain and artificial intelligence (AI). Blockchain can offer a decentralized and tamper-proof framework, ensuring the confidentiality and integrity of IoT data. AI can analyze large volumes of real-time data and detect anomalies in response to security threats in the IoT ecosystem. This paper explores the potential of these technologies and how they complement each other to provide a secured IoT system. Our main argument is that combining blockchain with AI can provide a robust solution for securing IoT networks and safeguarding the privacy of IoT users. This survey paper aims to provide a comprehensive understanding of the potential of these technologies for securing IoT networks and discuss the challenges and opportunities associated with their integration. It also provides a discussion on the current state of research on this topic and presents future research directions in this area.

Author 1: Md. Tauseef

Author 2: Manjunath R Kounte

Author 3: Abdul Haq Nalband

Author 4: Mohammed Riyaz Ahmed

Keywords: IoT; blockchain; AI; security; attacks; decentralization

Paper 99: Opportunities and Challenges in Human-Swarm Interaction: Systematic Review and Research Implications

Abstract: We conducted a Systematic Literature Review on scientific papers that examined the interaction between operators and drone swarms based on the use of a command and control center. We present the results of a meta-analysis of nine scientific papers published in the ACM DL and IEEE Xplore databases. Our findings show that research on human-drone swarm interaction shows a disproportionate interest in hand gestures compared to other input modalities for drone swarm control. Furthermore, all articles reviewed exclusively explored gestures and the size of the swarm used in the studies was limited, with a median of 3.0 and an average of 3.8 drones per study. We compiled an inventory of interaction modalities, recognition techniques, and application types from the scientific literature, which is presented in this paper. On the basis of our findings, we propose four areas for future research that can guide scientific investigations and practical developments in this field.

Author 1: Alexandru-Ionut Siean

Author 2: Bogdanel-Constantin Gradinaru

Author 3: Ovidiu-Ionut Gherman

Author 4: Mirela Danubianu

Author 5: Laurentiu-Dan Milici

Keywords: Human swarm interactions; input modalities; swarm control

Paper 100: IM2P-Medical: Towards Individual Management Privacy Preferences for the Medical Web Apps

Abstract: With the advancement of technology, people are now able to monitor their health more efficiently. Mobile phones and smartwatches are equipped with sensors that can measure real-time changes in blood pressure, SPO2, and other attributes and public them to service providers via web applications (called web apps) for health improvement suggestions. Moreover, users can share the collected health data with other people, such as doctors, relatives, or friends. However, using technology in healthcare has raised the issue of privacy. Some health web apps, by default, intrusively gather and share data. Additionally, smartwatches may monitor people’s health status 24/7. Therefore, users want to control how their health is processed (e.g., collected and shared). This can be cumbersome as they would have to configure each device manually. To address this problem, we have developed a privacy-preference prediction mechanism in the web apps called IM2P-Medical: towards Individual Management Privacy Preferences for the Medical web apps. To capture individual privacy preferences in the web apps, our model learns users’ privacy behavior based on their responses in different medical scenarios. In practice, we exploited several machine learning algorithms: SVM, Gradient Boosting Classifier, Ada Boost Classifier, and Gradient Boosting Regressor. To prove the effectiveness of the proposed model, we set up several scenarios to measure the accuracy as well as the satisfaction level in the two participant groups (i.e., expert and normal users). One key point in this research’s selection of participants is its focus on those living in developing countries, where privacy violation issues are not a common topic. The main contribution of our model is that it allows users to preserve their privacy without configuring privacy settings themselves.

Author 1: Nguyen Ngoc Phien

Author 2: Nguyen Thi Hoang Phuong

Author 3: Khiem G. Huynh

Author 4: Khanh H. Vo

Author 5: Phuc T. Nguyen

Author 6: Khoa D. Tran

Author 7: Bao Q. Tran

Author 8: Loc C. P. Van

Author 9: Duy T. Q. Nguyen

Author 10: Hieu M. Doan

Author 11: Bang K. Le

Author 12: Trong D. P. Nguyen

Author 13: Ngan T. K. Nguyen

Author 14: Huong H. Luong

Author 15: Duong Hon Minh

Keywords: Letter-of-credit; cash-on-delivery; blockchain; smart contract; NFT; ethereum; fantom; polygon; binance smart chain

Paper 101: Anomaly Discover: A New Community-based Approach for Detecting Anomalies in Social Networks

Abstract: In this paper, a new method called Anomaly Discover is provided for detecting anomalies in communities with mixed attributes (binary, numerical and categorical attributes). Our strategy tries to identify unusual users in Online Social Networks (OSN) communities and score them according to how far they deviate from typical users. Our ranking is based on both users’ attributes and network structure. Moreover, for effective anomaly detection, the context-selection process is performed for choosing relevant attributes that demonstrate a strong contrast between normal and abnormal users. So the anomaly score is defined as the degree of divergence in the network structure as well as a context-specific subset of attributes. To assess the efficacy of our model, we used real and artificial networks. We then compared the outcomes to those of two state-of-art models. The outcomes show that our model performs well since it outperforms other models and can pick up anomalies that competing models miss.

Author 1: Hedia Zardi

Author 2: Hajar Alrajhi

Keywords: Anomaly detection; community anomaly; anomaly ranking; social networks; relevant attributes

Paper 102: A Particle Swarm Optimization with Imbalance Initialization and Task Rescheduling for Task Offloading in Device-Edge-Cloud Computing

Abstract: Smart devices, e.g., smart-phones, internet-of-thing device, has been prevalent in our life. How to take full advantage of the limited resources to satisfy as many as requirements of users is still a challenge. Thus, in this paper, we focus on the task offloading problem to address the challenge by device-edge-cloud computing, by PSO improved with the imbalance initialization and the task scheduling. The imbalance initialization is to increase the probability that a task is assigned to a computing node such that the node provides a longer slack time. The task scheduling is to reassign tasks with deadline violations into other nodes, to improve the number of accepted tasks for each offloading solution. Extensive experiment results show that our proposed algorithm has better performance than other ten classical and up-to-data algorithms in both the maximization of the accepted task number, the resource utilization, as well as the processing rate.

Author 1: Hui Fu

Author 2: Guangyuan Li

Author 3: Fang Han

Author 4: Bo Wang

Keywords: Cloud computing; edge computing; task offloading; task scheduling; particle swarm optimization

Paper 103: Prediction of Air Quality and Pollution using Statistical Methods and Machine Learning Techniques

Abstract: Air pollution is a major environmental issue and machine learning techniques play an important role in analyzing and forecasting these data sets. Air quality is an outcome of the complex interaction of several factors involving the chemical reactions, meteorological parameters, and emissions from natural and anthropogenic sources. In this paper, we propose an efficient combined technique that takes the benefits of statistical techniques and machine learning techniques to predict/forecast the Air Quality and Pollution in particular regions. This work also indicates that prediction performance varies over different regions/cities in India. We used time series analysis, regression and Ada-boosting to anticipate PM 2.5 concentration levels in several locations throughout Hyderabad on an annual basis, depending on numerous atmospheric and surface parameters like wind speed, air temperature, pressure, and so on. Dataset for this investigation is taken from Kaggle and experimented with proposed method and comparison results of our experiments are then plotted.

Author 1: V. Devasekhar

Author 2: P. Natarajan

Keywords: Air quality; forecasting; machine learning; statistical techniques

Paper 104: Fraud Mitigation in Attendance Monitoring Systems using Dynamic QR Code, Geofencing and IMEI Technologies

Abstract: Attendance monitoring is a vital activity in several organizations. Due to its importance, many attendance monitoring systems have been developed to automate this process. Despite several advancements in automated attendance management solutions, attendance fraud remains an issue as some end users can manipulate known vulnerabilities, such as proxy attendance, buddy-punching, early departure, and so on. In this paper, a fraud-resistant attendance management solution is developed by harnessing technologies such as geofencing, dynamic QR code and IMEI Checking. The proposed solution is comprised of a single-page web application where QR code can be enabled for attendance registration, and a mobile application, where end-users can scan generated QR code to register their attendance. Attendance cheating via QR code sharing is prevented by en-coding the polygonal coordinates of the event venue in the QR code to determine if the user is within the venue. The proposed system solves the problem of proxy attendance by registering and verifying the end user’s device IMEI number. Results obtained from testing indicate that attempts at committing a variety of attendance frauds are effectively mitigated.

Author 1: Augustine Nwabuwe

Author 2: Baljinder Sanghera

Author 3: Temitope Alade

Author 4: Funminiyi Olajide

Keywords: Attendance management systems; fraud prevention; dynamic QR code; geofencing; IMEI verification; software algorithms; mobile application

Paper 105: Reverse Supply Chain Management Through a Quantity Flexibility Contract: A Case of Stochastic Remanufacturing Capacity

Abstract: This article investigates a two-echelon reverse supply chain (RSC) where a third-party logistics provider charges customers to return outdated products. A green manufacturer refurbishes qualified returned products through the remanufacturing process. Remanufacturing capacity is considered a stochastic variable. Under the volatility of remanufacturing capacity, some likely examined, and qualified products could not be remanufactured. If a collected product cannot be processed, it should be salvaged at a lower value and be perceived as a lost profit. In such scenarios, increasing the quantity of returned outdated products is suitable if there is a strong possibility of enough capacity in the remanufacturing process. This paper develops a stochastic model to identify the optimal order quantity under diverse contracts, including wholesale price, centralized, and quantity flexibility contracts. Under the quantity flexibility contract, the green supplier might cancel its preliminary order in a restricted quantity. Additionally, third-party logistics supplier offers a restricted quantity above the initial order to minimize understocking during peak seasons. Our numerical experiments demonstrate that the suggested quantity flexibility can coordinate the examined RSC under the volatility of remanufacturing capacity. Contrary to wholesale and centralized contracts, quantity flexibility is a more practical alternative from the perspective of participants’ profitability.

Author 1: Changhao Zhang

Keywords: Reverse supply chain; channel coordination; uncertain remanufacturing capacity; quantity flexibility contract

Paper 106: Simulation Method of Port Petrochemical Industry Throughput Development under the Background of Integration of Port, Industry and City

Abstract: In order to accurately predict the changes in the throughput of port petrochemical products and facilitate the formulation of relative decisions, this paper analyzes the factors affecting the throughput of port petrochemical products in a city through the GRA method. After sorting and selection, PCA method is used for pretreatment. In the SVM algorithm, ICSO is used to obtain the best parameters and improve the prediction accuracy and efficiency. In view of the variability of future development, three development scenarios are set up to prepare for the throughput forecast of petrochemical products in a city's port. The results show that the optimization speed of ICSO algorithm is very fast. When the training iteration is 20, the best fitness value is obtained, which is 0.0572. The training effect of ICSO-SVM algorithm is good, the gap between it and the original data is small, and the overall trend is close to the original data. In the test prediction, ICSO-SVM algorithm has the best prediction effect, and its MAE, RMSE and MAPE are the smallest. The minimum MAE is 762.2, 477.0 smaller than CSO-SVM algorithm, and the latter's MAE is 1239.2. The minimum MAPE of the proposed algorithm is 1.05%, while that of CSO-SVM algorithm is 1.71%. In general, the prediction error of ICSO-SVM algorithm is smaller. After the prediction of different development scenarios, the throughput of petrochemical products in a port of a city shows an increasing trend in the next five years. This method can be applied to the development forecast of port petrochemical products and provide reference for decision-making.

Author 1: Tingting Zhou

Author 2: Chen Guo

Keywords: Support vector machine algorithm; port throughput; chicken swarm optimization algorithm; grey correlation analysis; petrochemical products