IJACSA Volume 14 Issue 7 - thesai.org

IJACSA Volume 14 Issue 7

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: Secure Virtual Local Area Network Design and Implementation for Electronic Data Interchange

Abstract: Electronic Data Interchange is a popular platform for sensitive business transactional data transmission over the local to public network. It requires Value Added Network to communicate local data from one endpoint to another. In this paper, we proposed a value-added network design and deployment using the Virtual Local Area Network. The Value Added Network over a virtual Local Area Network eases the burden of managing the network and its functional devices. This proposed network provides the solution of data traffic as per demand required network path for the Electronic Data Interchange applications. This advanced deployment model of Value Added Network over a virtual Local Area Network also offers more robust security to the network entities and traffic. This proposed Virtual Local area network has been deployed with the Cisco environment, and all the specifications have been successfully implemented and tested for optimal security for Electronic Data Interchange. Virtual Local Area Network deployed with four different methods such as transferring packets with backbone, virtual Local Area Network using the tagging method, implementing virtual Local Area Network using Time Division Multiplexing, and by user-defined frame field. All these deployments are successfully done, and the secure platform for Electronic Data Interchange data exchange from one end local network to the other local network has been optimized.

Author 1: Jhansi Bharathi Madavarapu

Author 2: Firdous Hussain Mohammed

Author 3: Shailaja Salagrama

Author 4: Vimal Bibhu

Keywords: Electronic data interchange; value added network; virtual local area network; data link layer; time division multiplexing

Paper 2: An Intelligent Evaluation Path for English Teaching Quality: Construction of an Evaluation Model Based on Improved BPNN

Abstract: The current intelligent evaluation methods for English teaching quality are inefficient and have poor evaluation accuracy for effective assessment. The paper suggests an evaluation model based on an upgraded Back Propagation Neural Network to overcome the aforementioned issues. First, the principal component analysis is utilized to lessen the dimensionality of the index system as we build an English teaching quality evaluation index system with reference to the results of the existing study. Then, we adopt a multi-strategy improved dragonfly optimization algorithm to evaluate Back Propagation Neural Network for its defects; an algorithm to improve it. Finally, to increase the efficacy and objectivity of English teaching quality evaluation, an intelligent evaluation model based on IDA-BPNN is developed. The experimental results demonstrate that the IDA-BPNN model has an evaluation accuracy of 98.96%, an F1 value of 0.950 on the training set and 0.968 on the test set, a Recall value of 0.948 on the training set and 0.966 on the test set. The aforementioned indicators are all superior to the most recent state-of-the-art approaches for evaluating teaching quality. The aforementioned findings thus demonstrate that the model suggested in the study has high performance and can successfully improve the accuracy and efficiency of English teaching quality evaluation, which has a positive impact on the development of English teaching careers.

Author 1: Weihua Shen

Author 2: Wei Lu

Author 3: Yukun Qi

Keywords: Teaching quality evaluation; English; BPNN; intelligent evaluation; dragonfly optimization algorithm

Paper 3: Determinants of Medical Internet of Things Adoption in Healthcare and the Role of Demographic Factors Incorporating Modified UTAUT

Abstract: Medical Internet of Things (mIoT) is the IoT sub-set with vast potential in healthcare. However, the adoption of eHealth solutions such as mIoT has been a critical challenge in the health sector of the Kingdom of Saudi Arabia. Therefore, this study was conducted to explore the mIoT adoption determinants in Saudi public hospitals. Methods: A total of 271 participants were recruited from public hospitals in Riyadh, and a modified UTAUT model named UTAUT-HS was developed in this study to test its relevance with respect to mIoT adoption. Results: Ten path relationships were tested in this study, out of which six showed significant results. Similarly, three variables (Computer and English Language Self-efficacy or CESE, Performance Expectancy or PE and Social Influence or SI) showed a significant direct relationship with the behavioural intention to adopt mIoT. Furthermore, CESE showed the strongest relationship and emerged as a major sub-set of Effort Expectancy (EE) for mIoT adoption. However, moderator analysis showed substantial variations between different study demographic groups. In particular, the current study findings unravelled a comparatively novel relevance of Perceived Threat to Autonomy (PTA) for mIoT adoption for clinical and non-clinical and for older and younger participants. Conclusion: The study concludes that UTAUT-HS is an adequate model to explain the mIoT adoption in healthcare. However, it also suggests conducting future large-scale studies in KSA and elsewhere to validate the relevance of UTAUT-HS in other contexts and with much more confidence.

Author 1: Abdulaziz Alomari

Author 2: Ben Soh

Keywords: Medical internet of things; eHealth adoption; modified UTAUT; demographics and IT adoption

Paper 4: Apache Spark in Riot Games: A Case Study on Data Processing and Analytics

Abstract: This case study examines Riot Games' use of Apache Spark and its effects on data processing and analytics. Riot Games is a well-known game production studio. The developer Riot Games, best known for the well-liked online multiplayer game League of Legends, manages enormous volumes of data produced daily by millions of players. Riot Games handled and analyzed this data quickly using Apache Spark, a distributed computing technology that made insightful findings and improved user experiences. This case study explores Riot Games' difficulties, the company's adoption of Apache Spark, its implementation, and the advantages of utilizing Spark's capabilities. We evaluated the drawbacks and advantages of adopting Spark in the gaming sector and offered suggestions for game creators wishing to embrace Spark for their data processing and real-time analytics requirements. Our study adds to the increasing body of knowledge on the use of Spark in the gaming sector and offers suggestions and insights for both game producers and researchers.

Author 1: Kanhaiya Sharma

Author 2: Firdous Hussain Mohammad

Author 3: Deepak Parashar

Keywords: Riot games; Apache Spark; data processing; real-time analytics; distributed computing technology

Paper 5: Deep Learning-based Method for Enhancing the Detection of Arabic Authorship Attribution using Acoustic and Textual-based Features

Abstract: Authorship attribution (AA) is defined as the identification of the original author of an unseen text. It is found that the style of the author’s writing can change from one topic to another, but the author’s habits are still the same in different texts. The authorship attribution has been extensively studied for texts written in different languages such as English. However, few studies investigated the Arabic authorship attribution (AAA) due to the special challenges faced with the Arabic scripts. Additionally, there is a need to identify the authors of texts extracted from livestream broadcasting and the recorded speeches to protect the intellectual property of these authors. This paper aims to enhance the detection of Arabic authorship attribution by extracting different features and fusing the outputs of two deep learning models. The dataset used in this study was collected from the weekly livestream and recorded Arabic sermons that are available publicly on the official website of Al-Haramain in Saudi Arabia. The acoustic, textual and stylometric features were extracted for five authors. Then, the data were pre-processed and fed into the deep learning-based models (CNN architecture and its pre-trained ResNet34). After that the hard and soft voting ensemble methods were applied for combining the outputs of the applied models and improve the overall performance. The experimental results showed that the use of CNN with textual data obtained an acceptable performance using all evaluation metrics. Then, the performance of ResNet34 model with acoustic features outperformed the other models and obtained the accuracy of 90.34%. Finally, the results showed that the soft voting ensemble method enhanced the performance of AAA and outperformed the other method in terms of accuracy and precision, which obtained 93.19% and 0.9311 respectively.

Author 1: Mohammed Al-Sarem

Author 2: Faisal Saeed

Author 3: Sultan Noman Qasem

Author 4: Abdullah M Albarrak

Keywords: Authorship attribution; acoustic features; fusion approach; deep learning; CNN; ResNet34

Paper 6: Speech-Music Classification Model Based on Improved Neural Network and Beat Spectrum

Abstract: A speech-music classification method according to a developed neural system and beat spectrum is proposed to achieve accurate classification of speech-music through pre-emphasis, endpoint detection, framing, windowing and other steps to preprocess and collect vocal music signals. After fast Fourier transforms and triangle filter processing, the Mel frequency cepstrum coefficient (MFCC) is obtained, and a discrete cosine transform is performed to obtain the signal MFCC characteristic parameters. After calculating the similarity of feature parameters through cosine similarity, the signal similarity matrix is obtained, based on which the vocal music beat spectrum is obtained. The residual structure is optimized by adding Swish and max-out activation functions, respectively, between convolutional neural network layers to build residual convolution layers and deepen the number of convolution layers. The connected time series classification (CTC) is used as the objective loss function. It is applied to the softmax layer to build a deep optimization residual convolutional neural network for speech-music classification model. The pitch spectrum of vocal music is used as the input information of the model to realize the vocal music classification. The experiment proves that the classification accuracy of the design model is higher than 99%; when the iteration reaches 1200, the training loss approaches 0; when the signal-to-noise ratio is 180dB, the sensitivity and specificity are 99.98% and 99.96%, respectively; the accuracy of voice music classification is higher than 99%, and the running time is 0.48 seconds. It has been proven that the model has high classification accuracy, low training loss, good sensitivity and special effects, and can effectively achieve the classification of speech-music.

Author 1: Chun Huang

Author 2: Wei HeFu

Keywords: Vocal music; classification model; beat spectrum; feature parameter extraction; cosine similarity; convolutional neural network

Paper 7: A Method for Evaluating the Competitiveness of Human Resources in High-tech Enterprises Based on Self-organized Data Mining Algorithms

Abstract: The level of human resources competitiveness of high-tech companies affects the efficiency and effectiveness of enterprises to a particular extent. To achieve sustainable development of high-tech enterprises, an evaluation method of human resource competitiveness of high-tech enterprises based on a self-organized data mining algorithm is proposed. The fuzzy clustering algorithm is used to select five first-level indexes for the evaluation of HR competitiveness of high-tech companies, including human capital power, human resources policy incentive power, and human resources performance manifestation power, and to construct the initial evaluation indicator setting. The self-organized data mining algorithm is used to identify the key attributes related to the human resource competitiveness of high-tech companies within the initial assessment indicator setup, reduce the complexity of the indexes and construct the final rating index system. The multi-level fuzzy evaluation method is applied to calculate the evaluation index weights and fuzzy evaluation matrix to obtain the assessment results of HR competitivity of high-tech enterprises. The experimental results show that the information contribution rate of the evaluation index system constructed by this method is higher than 95%, which can accurately evaluate the human resource competitiveness of high-tech enterprises.

Author 1: Sun Zhixin

Keywords: Self-organized data mining algorithm; high-tech enterprises; human resources; competitiveness evaluation; multi-level fuzzy evaluation method

Paper 8: A Novel Internet of Things-enabled Approach to Monitor Patients’ Health Statistics

Abstract: Leveraging Internet of Things (IoT) technology in healthcare systems improves patient care, reduces costs, and increases efficiency. Enabled by IoT, telemedicine allows remote patient monitoring, vital sign tracking, and seamless data accessibility for doctors across multiple locations. This article presents a novel IoT-enabled approach that utilizes artificial neural networks with radial basis functions to detect patients' positions. This real-time tracking mechanism operates even without cellular connectivity, providing timely diagnoses and treatments. Our research aims to develop a smart and cost-effective healthcare approach, revolutionizing patient care. Mathematical analysis and experiments confirm the effectiveness of our proposed method, particularly in predicting patient location for the upcoming smart healthcare solution.

Author 1: Xi GOU

Keywords: Internet of things; healthcare; telemedicine; artificial neural network

Paper 9: Automated Modified Grey Wolf Optimizer for Identification of Unauthorized Requests in Software-defined Networks

Abstract: Software Defined Networking (SDN) is utilized to centralize network control within a controller, but its reliance on a single control plane can make it vulnerable to attacks such as DDoS. This highlights the importance of developing effective security mechanisms and using proactive measures such as detection and prevention strategies to mitigate the risk of attacks. Many DDoS attack detection technologies within SDN focus on detecting and mitigating the attack once it has occurred in the controller, which leads to more seconds of exposure, diminished precision, and high overhead. In this work, we have developed an Automated Modified Grey Wolf Optimizer Algorithm (AMGWOA) to design the detection of this malicious activity in an SDN environment to prevent the attack in the controller. Our methodology involves the development of the AMGWOA, which incorporates a mechanism to facilitate the blocking of malicious requests while reducing detection time and minimizing the use of storage and data resources for detection purposes. The results obtained show that our model performs well, with an ability to minimize a very large number of malicious requests in a minimum of time of less than 1 second compared to Grey Wolf Optimizer and particle swarm optimization algorithms evaluated using the same datasets.

Author 1: Aminata Dembele

Author 2: Elijah Mwangi

Author 3: Abderrahim Bouchair

Author 4: Kennedy K Ronoh

Author 5: Edwin O Ataro

Keywords: Software-defined networks; security; DDoS attacks; metaheuristic algorithms; Grey Wolf Optimizer

Paper 10: A Bibliometric Analysis of Research on Risks in the Poultry Farming Industry: Trends, Themes, Collaborations, and Technology Utilization

Abstract: This paper explores the risks prevalent in the poultry farming industry, drawing upon an extensive examination conducted by researchers over the past decade. Employing a bibliometric analysis approach, a comprehensive search of the Scopus database was conducted using relevant keywords related to poultry farming risk and technology utilization. The search spanned from 2002 to 2022, yielding 345 pertinent documents. This study presents an overview of the current state of publications concerning poultry farming risk and its intersection with technology utilization. It delves into citation patterns, prevalent themes, and authorship analysis, focusing on the role of technology in mitigating risks. The comprehensive citation analysis highlights the impact of technology-related studies in the field. Frequency analysis employed Microsoft Excel, while VOSviewer facilitated data visualization. Harzing's Publish or Perish software was used for citation metrics and analysis. The findings reveal a consistent increase in publications on risk in poultry farming since 2002, particularly in relation to technology utilization. The United States emerges as the most active country in this area of research, with Wageningen University from the Netherlands identified as the most prolific institution contributing significantly to risk in poultry farming research, including technology applications. The research involved 32 scholars from 70 different countries and 32 distinct institutions, reflecting the multi-authorship and multicultural nature of the research. It is important to note that this paper focuses solely on the Scopus database, while future researchers may consider alternative databases for new studies, recognizing the expanding role of technology in addressing risks in the poultry farming industry.

Author 1: Kamal Imran Mohd Sharif

Author 2: Mazni Omar

Author 3: Muhammad Danial Mohd Noor

Author 4: Mohd Azril Ismail

Author 5: Mohamad Ghozali Hassan

Author 6: Abdul Rehman Gilal

Keywords: Poultry farming risk; poultry farming industry; bibliometric analysis; Harzing’s Publish or Perish; VOSviewer

Paper 11: Adaptive Style Transfer Method of Art Works Based on Laplace Operator

Abstract: In order to improve the image quality of artworks after style transfer, the adaptive style transfer method of artworks based on the Laplace operator is studied. Through three steps of expansion processing, corrosion processing and multi-scale morphological enhancement, the image edge of the content of artworks is enhanced. The colour and brightness of the artworks with edge enhancement are transferred, and the transfer results are input into the convolution neural network simultaneously with the style image. According to the improved Laplace operator, the Laplace operator loss term of the convolution neural system is counted, the style losing term of the style picture of the art image is determined, and the total loss function is constructed. According to the determined loss function, a convolution neural network is used to output paintings' adaptive style transfer results. The experiential outcomes indicate that this technique is able to realize the adaptive style transmission of paintings. After style transfer, the picture quality of paintings is high, and the adaptive transfer of artworks can be realized within 500ms.

Author 1: HaiTing Jia

Keywords: Laplace operator; artworks; adaptive style transfer; brightness migration; convolution neural network

Paper 12: Application of Virtual Reality Technology in the Design of Interactive Interfaces for Public Service Announcements

Abstract: With the development of technology, more and more public service announcements are being designed with interactive interfaces. There are many different ways to interact with interactive interfaces, and using appropriate design methods can expand the impact of PSAs. The study incorporates image pre-processing methods based on virtual reality, using the cvtColor grey scale function and median filtering method to process the images, an iterative approach to camera positioning method design, and subsequent performance testing of the research algorithms. The test results showed that the peak signal-to-noise ratio of the research method was 13.390dB in the image pre-processing process and 35.635dB on lightly shaded images; in the error test, the rotational mean error of the research method was approximately 4.2degrees at four reference points; and in the image plane reprojection test, 70% of the points of the research method almost coincided with the original point position. The method generated 1203 designs with 40 reference points. The experimental results show that the research method can effectively design interactive interfaces for PSAs in virtual reality environments, and can propose more design solutions, and has better performance in virtual reality environment positioning.

Author 1: Rong Hu

Keywords: Virtual reality; interactive interfaces; interface design; greyscaling; image pre-processing

Paper 13: Method for Image Quality Evaluation of Satellite-based SAR Data

Abstract: A method for image quality evaluation of satellite-based Synthetic Aperture Radar: SAR data is proposed. Not only geometric fidelity but also signal to noise ratio, frequency component, saturated pixel ratio, speckle noise, optimum filter kernel size and its filter function are evaluated. Through experiments with SAR so called QPS-SAR_2 (Q-shu Pioneers of Space SAR the second) of imagery data, all these items are evaluated, and it is confirmed that the geometric and radiometric performances are good enough. Also, geometric fidelity of QPS-SAR_2 is compared to Sentinel-1/SAR European Space Agency (ESA) provided data which is obtained on the same day of QPS-SAR_2 data acquisition.

Author 1: Kohei Arai

Author 2: Michihiro Mikamo

Author 3: Shunsuke Onishi

Keywords: Image quality; synthetic aperture radar (SAR); geometric fidelity; signal to noise ratio; frequency component; saturated pixel ratio; speckle noise; optimum filter kernel size; filter function for speckle noise reduction

Paper 14: A Multi-label Filter Feature Selection Method Based on Approximate Pareto Dominance

Abstract: The Pareto dominance has been applied to resolve the issue of choosing significant features from a multi-label dataset. High-dimensional labels will directly result in the difficulty of forming Pareto dominance. This work proposes a multi-label feature selection approach based on the approximate Pareto dominance (MAPD) to address this issue. It maps the multi-label feature selection to the problem of solving the approximate Pareto dominant solution set. By introducing an approximate parameter, it is possible to efficiently cut down on the amount of features in the chosen feature subset while also raising its quality. To verify the performance of MAPD, this research compares the MAPD algorithm with alternative approaches in terms of Hamming loss, accuracy, and chosen feature size using nine publicly available multi-label datasets. The findings indicate that the MAPD method performs better in terms of classification accuracy, Hamming loss, and the amount of features that may be chosen.

Author 1: Jian Zhou

Author 2: Yinnong Guo

Keywords: Approximate Pareto dominance; multi-label data; feature selection

Paper 15: Dynamic Allocation Method of Incentive Pool for Financial Management Teaching Innovation Team Based on Data Mining

Abstract: In order to reasonably allocate the amount of incentive pool and promote the unity of members of the financial management teaching innovation team, a dynamic allocation method of incentive pool for the financial management teaching innovation team based on data mining is proposed. This method constructs the incentive pool allocation index system by analyzing the principles of risk and income correlation, income and contribution consistency, individual and overall profit consistency, as well as the actual contribution of the financial management teaching innovation team, members' efforts and other factors that affect the allocation of incentive pool. After determining the index weight, the maximum entropy model is used to establish the incentive pool function of the financial management teaching innovation team project. The incentive pool scale decision model is established according to the prospect theory. After outputting the scale of the financial management teaching innovation team's incentive pool using the construction model, the incentive pool model of the financial management teaching innovation team is obtained. Based on the asymmetric Nash negotiation model, the allocation model for the incentive pool model of the financial management teaching innovation team is established, the improved artificial colony algorithm in the data mining algorithm is used to solve the model, and the dynamic allocation result of the incentive pool of the financial management teaching innovation team is obtained. The experiment shows that this method can effectively calculate the size of the incentive pool and allocate the incentive pool. The members of the financial management teaching innovation team have a high degree of satisfaction with the allocation result of the incentive pool, with allocation satisfaction consistently fluctuating around 96%.

Author 1: Huang Jingjing

Author 2: Zhang Xu

Keywords: Data mining; financial management; teaching innovation team; incentive pool; dynamic allocation; artificial colony

Paper 16: A Hybrid Federated Learning Framework and Multi-Party Communication for Cyber-Security Analysis

Abstract: The term "Internet of Things" (IoT) describes a global system of electronically linked devices and sensors capable of two-way communication and data sharing. IoT provides various advantages, including improved efficiency and production and lower operating expenses. Concern about data breaches is constantly present, for example, since devices with sensors capture and send confidential data that might have dire effects if leaked. Hence, this research proposed a novel hybrid federated learning framework with multi-party communication (FLbMPC) to address the cyber-security challenges. The proposed approach comprises four phases: data collection and standardization, model training, data aggregation, and attack detection. The research uses the UNSW-NB15 cyber-security dataset, which was collected and standardized using the z-score normalization approach. Federated learning was used to train the local models of each IoT device with their respective subsets of data. The MPC method is used to aggregate the encrypted local models into a global model while maintaining the confidentiality of the local models. Finally, in the attack detection phase, the global model compares real-time sensor data and predicted values to identify cyber-attacks. The experiment findings show that the suggested model outperforms the current methods in terms of accuracy, precision, f-measure and recall.

Author 1: Fahad Alqurashi

Keywords: Federated learning; multi-party communication; cyber-security; machine learning; internet of things

Paper 17: Explainable Artificial Intelligence (XAI) for the Prediction of Diabetes Management: An Ensemble Approach

Abstract: Machine learning determines patterns from data to expedite the process of decision making. Fact-based decisions and data-driven decisions are specified by the industry specialist. Due to the continuous growth of machine language models in healthcare, they are breeding continuous complexity and black boxes in ML models. To make the ML model crystal clear and authentically explainable, AI accession came in prevalence. This research scrutinizes the explainable AI and capabilities in the Indian healthcare system to detect diabetes. LIME and SHAP are two libraries and packages that are used to implement explainable AI. The intimated base amalgamates the local and global interpretable methods, which enhances the crystallinity of the complex model and obtains intuition into the equity from the complex model. Moreover, the obtained intuition could also boost clinical data scientists to plan a more felicitous composition of computer-aided diagnosis. Importance of XAI to forecast stubborn disease. In this case, of stubborn diabetes, the correlation between plasma versus insulin, age versus pregnancies, class (diabetic and nondiabetic) versus plasma glucose persisted with a strong relationship. The PIDD (PIMA Indian Diabetic Data set) with the SHAP value is used for concise dependency, and LIME is applicable when anchors and importance of features are both required simultaneously. Dependency plots help physicians visualize independent relationships with predicted disease. To identify dependencies of different attributes, a correlation heatmap is used. From an academic perspective, XAI is very indispensable to mature in the near future. To estimate the presentation of other applicable data set correspondence studies are very much apprenticed.

Author 1: Rita Ganguly

Author 2: Dharmpal Singh

Keywords: Explainable Artificial Intelligence (XAI); diabetes; interpretability; machine learning; chronic disease management

Paper 18: Research on the Text Classification of Legal Consultation Based on Deep Learning

Abstract: In view of the existing traditional legal service, practitioners are unable to meet the huge demand; a large number of citizens are unable to determine the scope of the problems when they encounter infringement or require various legal assistance. Based on this, an automatic classification model of legal consultation based on Deep Learning is proposed in this paper. A KP+BiLSTM+Attention model is proposed. The Keyword Parser is introduced to extract key information. TF-IDF and part of speech tagging are used to filter out the important information in the user's legal problem description. The extracted keywords are given a weight value, and the other information weights are set to zero. The text information is transferred into two parallel word vector embedding layers. One of the word vector embedding layers transfers the results to the fusion layer for splicing, difference and point multiplication after the key information is converted into vector form. The output results are respectively connected with the results obtained from the other embedding layer as residuals. The final results are transferred to the BiLSTM+Attention model for training. The test results show that KP+BiLSTM+Attention model has significantly improved the accuracy and F1 value of the best benchmark method for text classification tasks of legal consulting. Therefore, KP+BiLSTM+Attention method has better performance in dealing with the classification of legal consulting issues.

Author 1: ZuoQiang Du

Keywords: Text classification; legal consultation; deep learning; KP+BILSTM+ATT model; word embedding layer

Paper 19: Continuous Software Engineering for Augmented Reality

Abstract: Continuous software engineering is a new trend that has attracted increasing attention from the research community in recent years. In software engineering there are “continuous” stages that are used depending on the number of artifact repositories such as databases, meta data, virtual machines, networks and servers, various logs, and reports. Augmented Reality (AR) technology is currently growing rapidly. We can find this technology in various fields of life, but unfortunately sustainable software engineering for Augmented Reality is not found. The method shown in previous research is a general method in software engineering so that a theory is needed for sustainable software engineering for AR considering that AR is not just an ordinary application but there are 3D elements and specific components that must be met so that it can be called AR. The main idea behind this research is to find a continuous pattern from the stages of the existing method so far. For example, in general the stages of system development are planning, analysis, design, implementation and maintenance. Then after the application has been built, does it finish there? As we know software always grows and develops according to human needs. Therefore, there are continuous stages that must be patterned so that the life cycle process can be maintained. In this paper we present our initial findings about the continuous stages of continuous software engineering namely continuous planning, continuous analysis, continuous design, continuous programming, continuous integration, and continuous maintenance.

Author 1: Suzanna

Author 2: Sasmoko

Author 3: Ford Lumban Gaol

Author 4: Tanty Oktavia

Keywords: Continuous software engineering; augmented reality; method in software engineering; continuous planning; continuous analysis; continuous design; continuous programming; continuous integration; continuous maintenance

Paper 20: A Local-Global Graph Convolutional Network for Depression Recognition using EEG Signals

Abstract: Graph Convolutional Networks (GCNs) have shown remarkable capabilities in learning the topological relationships among electroencephalogram (EEG) channels for recognizing depression. However, existing GCN methods often focus on a single spatial pattern, disregarding the relevant connectivity of local functional regions and neglecting the data dependency of the original EEG data. To address these limitations, we introduce the Local-Global GCN (LG-GCN), a novel GCN inspired by brain science research, which learns the local-global graph representation of EEG. Our approach leverages discriminative features extracted from EEG signals as auxiliary information to capture dynamic multi-level spatial information between EEG channels. Specifically, the representation learning of the topological space in brain regions comprises two graphs: one for exploring augmentation information in local functional regions and another for extracting global dynamic information. The aggregation of multiple graphs enables the GCN to acquire more robust features. Additionally, we develop an Information Enhancement Module (IEM) to capture multi-dimensional fused features. Extensive experiments conducted on public datasets demonstrate that our proposed method surpasses state-of-the-art (SOTA) models, achieving an impressive accuracy of 99.30% in depression recognition.

Author 1: Yu Chen

Author 2: Xiuxiu Hu

Author 3: Lihua Xia

Keywords: Electroencephalogram; depression recognition; Local-Global Graph Convolutional Network (LG-GCN); multilevel spatial information; brain regions; multiple graphs

Paper 21: University’s Service Delivery Improvement Through a DSS-enabled Client Feedback System

Abstract: The expansion of products and services on a global scale demands the improvement of an organization’s performance. In addition to addressing the challenges of improving product and service delivery, companies must focus not only on meeting customer expectations but also on surpassing them. Consequently, valuing the opinions of clients, giving the best client experience, and measuring client satisfaction are deemed vital not only for the company’s survival but also for gaining a competitive edge for the organizations in the wired communities. It is because of these premises that the Client Feedback System was developed in this study for the university’s service delivery improvement. This system captured the results of the Client Satisfaction Survey for School Year 2015-2016 to School Year 2020-2021. Interpretation of these captured data were made and action for the improvement of service delivery for each department in this university was recommended using the Decision Support System (DSS) technique. The system was created using the Rapid Application Development (RAD) method and utilized various software and technologies such as HTML, CSS, and JavaScript for the front-end development, MySQL and PHP for the back-end, and Apache as the local server of the system during its development and pilot testing.

Author 1: Belen M. Tapado

Author 2: John Gregory M. Bola

Author 3: Erickson T. Salazar

Author 4: Zcel T. Tablizo

Keywords: Decision support system; client satisfaction; client feedback system; rapid application development; service delivery improvement

Paper 22: Development of a Two-dimensional Animation for Business Law: Elements of a Valid Contract

Abstract: Elements of a valid contract is an important topic in corporate law. Since there are so many elements and related case studies, some students have difficulty remembering all the elements. Therefore, an animation will contribute to explaining all the elements in simpler terms, help the students remember the relevant case studies, and help lecturers teach students in an easier and more interactive way. An investigation on a 2D animation design and its effectiveness for corporate law and commerce learning is presented in this article. This paper aims to examine the 2D animation principle in animated explainer videos. In addition, the objective of this research is to develop an animation and evaluate the effectiveness of the 2D animation for Business Law teaching and learning. A comprehensive analysis of the 2D animation used in Business Law learning which focusing on spreading the importance of student understanding and motivation in Business Law course using a 2D animated approach is the expected outcome of this paper. The project collaborates with the Department of Commerce, Politeknik Melaka, Malaysia, for content expertise and testing. The Multimedia Production Process is the methodology used for the development of this research work, and the ADDIE Model is applied for the instructional design. The application is developed using Adobe After Effects, Adobe Premiere Pro, Adobe Media Encoder, and the Audacity platform. The contribution of this study is obvious, as the resulting outcomes can be used as guidelines for best practises of learning styles. The implications of this study will impact teaching and learning and increase understanding. This research is expected to improve teaching delivery while also increasing user understanding and motivation to learn.

Author 1: Sarni Suhaila Rahim

Author 2: Hazira Saleh

Author 3: Nur Zulaiha Fadlan Faizal

Author 4: Shahril Parumo

Keywords: 2D animation; business law; elements of a valid contract; teaching and learning; multimedia

Paper 23: An Improved Lane-Keeping Controller for Autonomous Vehicles Leveraging an Integrated CNN-LSTM Approach

Abstract: Representing the task of navigating a car through traffic using traditional algorithms is a complex endeavor that presents significant challenges. To overcome this, researchers have started training artificial neural networks using data from front-facing cameras, combined with corresponding steering angles. However, many current solutions focus solely on the visual information from the camera frames, overlooking the important temporal relationships between these frames. This paper introduces a novel approach to end-to-end steering control by combining a VGG16 convolutional neural network (CNN) architecture with Long Short-Term Memory (LSTM). This integrated model enables the learning of both the temporal dependencies within a sequence of images and the dynamics of the control process. Furthermore, we will present and evaluate the estimated accuracy of the proposed approach for steering angle prediction, comparing it with various CNN models including the Nvidia classic model, Nvidia model, and MobilenetV2 model when integrated with LSTM. The proposed method demonstrates superior accuracy compared to other approaches, achieving the lowest loss function. To evaluate its performance, we recorded a video and saved the corresponding steering angle results based on human perception from the robot operating system (ROS2). The videos are then split into image sequences to be smoothly fed into the processing model for training.

Author 1: Hoang Tran Ngoc

Author 2: Phuc Phan Hong

Author 3: Nghi Nguyen Vinh

Author 4: Nguyen Nguyen Trung

Author 5: Khang Hoang Nguyen

Author 6: Luyl-Da Quach

Keywords: End-to-end steering control; convolutional neural network; LSTM; nvidia model; MobileNetv2; VGG16

Paper 24: Lane Road Segmentation Based on Improved UNet Architecture for Autonomous Driving

Abstract: This paper introduces a real-time workflow for implementing neural networks in the context of autonomous driving. The UNet architecture is specifically selected for road segmentation due to its strong performance and low complexity. To further improve the model's capabilities, Local Binary Convolution (LBC) is incorporated into the skip connections, enhancing feature extraction, and elevating the Intersection over Union (IoU) metric. The performance evaluation of the model focuses on road detection, utilizing the IOU metric. Two datasets are used for training and validation: the widely used KITTI dataset and a custom dataset collected within the ROS2 environment. Simulation validation is performed on both datasets to assess the performance of our model. The evaluation of our model on the KITTI dataset demonstrates an impressive IoU score of 97.90% for road segmentation. Moreover, when evaluated on our custom dataset, our model achieves an IoU score of 98.88%, which is comparable to the performance of conventional UNet models. Our proposed method to reconstruct the model structure and provide input feature extraction can effectively improve the performance of existing lane road segmentation methods.

Author 1: Hoang Tran Ngoc

Author 2: Huynh Vu Nhu Nguyen

Author 3: Khang Hoang Nguyen

Author 4: Luyl-Da Quach

Keywords: Local binary patterns; feature extraction; UNet; semantic segmentation

Paper 25: Leveraging Big Data and AI in Mobile Shopping: A Study in the Context of Jordan

Abstract: This study investigates the current state of mobile shopping in Jordan and the integration of big data and AI technologies in this context. A mixed-methods approach, combining qualitative and quantitative data collection techniques, utilized to gather comprehensive insights. The survey questionnaire distributed to 105 individuals engaged in mobile shopping in Jordan. The findings highlight the popularity of mobile shopping and the preference for mobile apps as the primary platform. Personalized product recommendations emerged as a crucial factor in enhancing the mobile shopping experience. Privacy concerns regarding data sharing were present among respondents. Trust in AI-powered virtual assistants varied, indicating the potential for leveraging AI technologies. Respondents recognized the potential of big data and AI in improving the mobile shopping experience. The study concludes that businesses can enhance mobile shopping by utilizing AI-powered virtual assistants and prioritizing data security. The findings contribute to understanding mobile shopping dynamics and provide guidance for businesses and policymakers in optimizing mobile shopping experiences and driving economic growth in Jordan's digital economy. Future research and implementation efforts are encouraged to harness the potential of big data and AI in the mobile shopping landscape.

Author 1: Maher Abuhamdeh

Author 2: Osama Qtaish

Author 3: Hasan Kanaker

Author 4: Ahmad Alshanty

Author 5: Nidal Yousef

Author 6: Abdulla Mousa AlAli

Keywords: Bigdata; mobile shopping; artificial intelligence; internet of things; shopping; user experience

Paper 26: Optimizing Drying Efficiency Through an IoT-based Direct Solar Dryer System: Integration of Web Data Logger and SMS Notification

Abstract: Various agricultural and culinary products are dried to extend their shelf lives, mostly for marine foods. In many coastal locations of the Philippines, drying fish traditionally is still practiced, although study has shown that due to weather conditions and other factors, this technique is not seen to be reliable or cost-effective. The Internet of Things (IoT)-based Direct Solar Dryer System is optimized for drying efficiency by combining a web data logger and SMS notification system using Arduino Uno and ESP-32 to address difficulties with reliability and cost effectiveness. The study focuses on the potential and system efficiency of drying Sardinella fish (Tamban) in Brgy, Calibunan, Agusan Del Norte, Philippines; as well as investigating and assessing temperature, heat index, humidity, and temperature range alert conditions using a web application portal to serve as a remote monitoring platform for dependable data visualizations. The system delivered the expected results because the direct solar drier was able to raise and maintain the requisite temperature to accelerate drying while keeping the acceptable relative humidity. Furthermore, the system's monitoring and notification capabilities, as well as effective data collecting and data display via physical and remote monitoring, are supported by SMS notifications. As a result, the effectiveness of upgrading traditional sun drying with IoT technology can help reduce the challenges and disadvantages that fish drying farmers have faced. The study, with correct drying monitoring criteria, could serve as a model for other food products that can be dried.

Author 1: Joel I. Miano

Author 2: Michael A. Nabua

Author 3: Alexander R. Gaw

Author 4: Apple Rose B. Alce

Author 5: Cris Argie M. Ecleo

Author 6: Jewelane V. Repulle

Author 7: Jaafar J. Omar

Keywords: Arduino Uno; Internet of Thing (IoT); solar dryer system; web application portal; ESP-32; SMS notification

Paper 27: Anti-Spoofing in Medical Employee's Email using Machine Learning Uclassify Algorithm

Abstract: Since the advent of COVID-19, healthcare and IT cybersecurity have been an issue. Digital services and foreign labor have increased cyberattacks. July 2021 saw 260,642 phishing emails. 94% of 12 countries’ employees experienced epidemic cyberattacks. Phishing attacks steal sensitive data from spam emails or legitimate websites for profit. Phishing spam uses URL, domain, page, and content variables. Simple machine-learning methods stop phishing emails. This study discusses phishing emails and patient data and healthcare employee accounts cybersecurity. This paper covers COVID-19 email and phishing detection. This article examines the message's URL, subject, email, and links. Uclassify classifies content, spam, and languages and automates emails. Semi-supervised machine learning dominates healthcare. The Uclassify algorithm used multinomial Naive Bayesian classifiers. Document class is [0–1]. This article compared Multinomial Naive Bayesian in two experiments with other algorithms. Experiment 1 achieved an MNB accuracy of 96% based on a database from Kaggle Phishing. Experiment 2 showed that the Multinomial Naive Bayesian system accurately predicted URL and hyperlink targets based on PhishTank data. 96.67% of respondents correctly identified URLs, and 91.6% did so for hyperlinks. These two experiments focused on Tokenization, Lemmatization, and Feature Extraction (FE) and contained an internal feature set (IFS) and an external feature set (EFS). MNB is more exact than earlier methods since it uses decimal digits and word frequency. MNB only takes binary inputs. MNB can detect phishing and spoofing.

Author 1: Bander Nasser Almousa

Author 2: Diaa Mohammed Uliyan

Keywords: Spoofing; phishing; machine learning; Uclassify algorithm; medical employee's email

Paper 28: Adaptive Visual Sentiment Prediction Model Based on Event Concepts and Object Detection Techniques in Social Media

Abstract: Now-a-days, the increasing number of smartphones has caused the immediate sharing of photographs capturing current events on social media. The sentimental content of pictures from social events starts to be obtained from visual material, so visual sentiment analysis is a vital research topic. The research aims to reach valuable criteria to modify the visual sentiment prediction model based on event concepts and object detection techniques. In addition to adapting the approach for designing the method for predicting visual sentiments in a social network according to concept scores and measuring the performance of the model for predicting visual sentiments as accurately as possible, approach obtains a visual summary of social event images based on the visual elements that appear in the pictures which exceed sentiment-specific features. By this method, attributes (color, texture) are assigned to sentiments with discovering affective objects that are used to obtain emotions related to a picture of a social event by mapping the top predicted qualities to feelings and extracting the prevailing emotion connected with a photograph of a social event. This method is valid for a wide range of social events. This strategy also demonstrates the social event's effectiveness for a difficult social event image collection by using techniques for classifying complicated event images into sentiments, whether positive or negative.

Author 1: Yasser Fouad

Author 2: Ahmed M. Osman

Author 3: Samah A. Z. Hassan

Author 4: Hazem M. El-Bakry

Author 5: Ahmed M. Elshewey

Keywords: Sentiment Analysis (SA); visual sentiment analysis; image analysis; object recognition; event concepts; events concepts with object detection

Paper 29: Attacks on the Vehicle Ad-hoc Network from Cyberspace

Abstract: The emergence of Vehicle Ad hoc Networks (VANET) in 2003 has brought about a significant advancement in mobile phone networks and VANETs enable cars on the road to communicate with each other and the infrastructure on the street through a set of sensors and Intelligent Transport Systems (ITS). However VANETs are a low-level trust environment, making them vulnerable to misbehavior attacks and abnormal use. Thus, it is crucial to ensure that VANET systems and applications are secure and protected from cyber-attacks. This research aims to identify security challenges and vulnerabilities in VANET and proposes an algorithm that checks vehicle identity, location, and speed to detect and classify suspicious behavior. The research involves a study of the structures, architecture, and applications using VANET technology, the interconnection processes between them, and the types of architecture, layers, and applications that can pose a high risk. The research also focuses on the Confidentiality, Integrity and Availability (CIA) information security triangle and develops a program that uses machine learning to classify and analyze risks, attacks. The proposed algorithm provides security and safety for everyone on the road by identifying harmful behaviors of vehicles through knowledge of their location and identity. Overall, this research contributes to the development of a stable and secure Vehicular ad hoc network environment, enabling the integration of VANET security with smart cities.

Author 1: Anas Alwasel

Author 2: Shailendra Mishra

Author 3: Mohammed AlShehri

Keywords: Vehicular Ad hoc Network (VANET); Mobile Ad hoc Network (MANET); machine learning; random forest; linear regression

Paper 30: Optimization Solutions for Solving Travelling Salesman Problem in Graph Theory using African Buffalo Mechanism

Abstract: The African Buffalo Optimization (ABO), a metaheuristic optimization algorithm created from thorough study of African buffalos, a species of African cows, in African woods and savannahs, is suggested in this study. In its pursuit for food across the African continent, this animal demonstrates unusual intelligence, sophisticated organising capabilities, and remarkable navigational acumen. The African Buffalo Optimization creates a mathematical model based on this animal's behaviour and uses it to tackle several benchmark symmetrical Travel Salesman's Problem and six tough asymmetric Travelling Salesman Problem Library (TSPLIB) instances. Buffalos can ensure the effective exploitation and exploration of the problem space by frequent contact, teamwork, and a sharp mind of previous record discoveries, as well as tapping into the breed's collective exploits, according to this study. The results produced by solving these TSP problems using the ABO were compared to those obtained by utilizing other prominent methods. The results indicate that ABO gently outperformed than Lin-Kernighan and HBMO optimising solutions to the ATSP cases under investigative process, with a slightly higher accuracy of 99.5% compared to 87% for Lin-Kernighan and 80% for HBMO. The African Buffalo Optimization algorithm produces very competitive outcomes.

Author 1: Yousef Methkal Abd Algani

Keywords: African buffalo optimization; solutions; travelling salesman’s problem; graph theory

Paper 31: Multi-feature Fusion for Relation Extraction using Entity Types and Word Dependencies

Abstract: Most existing methods do not make full use of different types of information sources to extract effective features for relation extraction. This paper proposes a multi-feature fusion model based on raw input sentences and external knowledge sources, which deeply integrates diverse lexical, semantic, and syntactic features into deep neural network models. Specifically, our model extracts lexical features of different granularity from the original input text representation, entity type features from the entity annotation information of the corpus, and dependency features from the dependency trees. Meanwhile, the dimension-based attention mechanism is proposed to enrich the diversity of entity type features and enhance their discriminability. Different features enable the model to comprehensively utilize various types of information, so this paper fuses these features and train a classifier for relation extraction. The experimental results show that the proposed model outperforms the existing state-of-the-art baselines on the TACRED Revisited, Re-TACRED, and SemEval datasets, with macro-average F1 scores of 81.2%, 90.2%, and 89.4%, respectively, improving the performance by 1.4%, 4.4%, and 2% on average, which indicates the effectiveness of multi-feature fusion modeling.

Author 1: Pu Zhang

Author 2: Junwei Li

Author 3: Sixing Chen

Author 4: Jingyu Zhang

Author 5: Libo Tang

Keywords: Relation extraction; multi-feature fusion; information extraction; dependency tree; entity type

Paper 32: Mobile Apps Performance Testing as a Service for Parallel Test Execution and Automatic Test Result Analysis

Abstract: Now-a-days, numerous mobile apps are developed daily that influence the lives of people worldwide. Mobile apps are implemented within a limited time and budget. This is to keep up with the rapid business growth and to gain a competitive advantage in the market. Performance testing is a crucial activity that evaluates the behavior of the application under test (AUT) under various workloads. Performance testing in the domain of mobile app development is still a manual and time-consuming activity. As a negative consequence, performance testing is ignored during the development of many mobile apps. Thus, mobile apps may suffer from weak performance that badly affects the user experience. Therefore, cloud technology is introduced as a solution that emerges in the domain of software testing. Based on this technology, software testing is provided as a service (TaaS) that leverages cloud-based resources. This overcomes the testing issues and achieves high test quality. In this paper, a cloud-based testing as a service architecture is proposed for performance testing of mobile apps. The proposed performance testing as a service (P-TaaS) adopts efficient approaches for automating the entire process. Efficient approaches for test case generation, parallel test execution, and test results analysis are introduced. The proposed test case generation approach applies model-based testing (MBT) technique that generates test cases automatically from the AUT’s specification models and artifacts. The proposed P-TaaS lessens the testing time and satisfies the fast time-to-release constraint of mobile apps. Additionally, the proposed P-TaaS maximizes resource utilization, and allows continuous resource monitoring.

Author 1: Amira Ali

Author 2: Huda Amin Maghawry

Author 3: Nagwa Badr

Keywords: Performance testing; mobile apps testing; mobile apps performance testing; automated testing; cloud computing; TaaS; model-based testing

Paper 33: A Segmentation-based Token Identification for Recognition of Audio Mathematical Expression

Abstract: In human-computer interaction, humans can interact with the computer with the help of text, audio, images, speech, etc. Interacting with the computer using speech, speech recognitions in particularly audio segmentation is a challenging task due to accent or way of pronouncing style. To input mathematical symbols, words, functions, and expressions with the help of a keyboard are tedious and time-consuming. Input this with the help of audio, speeds up the input process. In this paper, an SBTI (audio Segmentation Based Token Identification) algorithm is proposed for the recognition of words in an audio mathematical expression. 6 types of audio mathematical expressions are considered for recognition. The proposed algorithm segments the audio file into chunks and from each chunk temporal and spectral characteristics of audio signals are selected to extract the features. The model is trained using a neural network. The proposed algorithm shows a classification accuracy of 100% for the algebraic, quadratic, area, and differentiation expression, 99% for trigonometric expression, and 92% for summation expression.

Author 1: Vaishali A. Kherdekar

Author 2: Sachin A. Naik

Author 3: Prafulla Bafna

Keywords: Audio segmentation; classification; feature extraction; neural network; speech recognition

Paper 34: Evaluating Machine Learning Models for Predicting Graduation Timelines in Moroccan Universities

Abstract: The escalating student numbers in Moroccan universities have intensified the complexities of managing on-time graduation. In this context, Machine learning methodologies were utilized to analyze the patterns and predict on-time graduation rates in a comprehensive manner. Our dataset comprised information from 5236 bachelor students who graduated in the years 2020 and 2021 from the Faculty of Law, Economic, and Social Sciences at Moulay Ismail University. The dataset incorporated a diverse range of student attributes including age, marital status, gender, nationality, socio-economic category of parents, profession, disability status, province of residence, high school diploma attainment, and academic honors, all contributing to a comprehensive understanding of the factors influencing graduation outcomes. Implementation and evaluation of the performance of five different machine learning models: Support Vector Machines, Decision Tree, Naive Bayes, Logistic Regression, and Random Forest, were carried out. These models were assessed based on their classification reports, confusion matrices, and Receiver Operating Characteristic (ROC) curves. From the findings, the Random Forest model emerged as the most accurate in predicting on-time graduation, showcasing the highest accuracy and ROC AUC score. Despite these promising results, it is believed that performance enhancements can be achieved through further tuning and preprocessing of the dataset. Insights from this study could enable Moroccan universities, among others, to better comprehend the factors influencing on-time graduation and implement appropriate measures to improve academic outcomes.

Author 1: Azeddine Sadqui

Author 2: Merouane Ertel

Author 3: Hicham Sadiki

Author 4: Said Amali

Keywords: Machine learning; logistic regression; classification reports; on time graduation; Moroccan universities

Paper 35: Efficient and Accurate Beach Litter Detection Method Based on QSB-YOLO

Abstract: Because of the potential threats it presents to marine ecosystems and human health, beach litter is becoming a major global environmental issue. The traditional manual sampling survey of beach litter is poor in real-time, poor in effect, and limited in the detection area, so it is extremely difficult to quickly clean up and recycle beach litter. Deep learning technology is quickly advancing, opening up a new method for monitoring beach litter. A QSB-YOLO beach litter detection approach based on the improved YOLOv7 is proposed for the problem of missed and false detection in beach litter detection. First, YOLOv7 is combined with the quantization-friendly Quantization-Aware RepVGG (QARepVGG) to reduce the model's parameters while maintaining its performance advantage. Secondly, A Simple, Parameter-Free Attention Module (SimAM) is used in YOLOv7 to enhance the feature extraction capacity of the network for the image region of interest. Finally, improving the original neck by combining the concept of the Bidirectional Feature Pyramid Network (BiFPN) allows the network to better learn features of various sizes. The test results on the self-built dataset demonstrate that: (1) QSB-YOLO has a good detection effect for six types of beach litter; (2) QSB-YOLO has a 5.8% higher mAP compared to YOLOv7, with a 43% faster detection speed, and QSB-YOLO has the highest detection accuracy for styrofoam, plastic products, and paper products; (3) QSB-YOLO has the greatest detection accuracy and detection efficiency when comparing the detection effects in various models. The results of the experiments demonstrate that the suggested model satisfies the need for beach litter identification in real-time.

Author 1: Hanling Zhu

Author 2: Daoheng Zhu

Author 3: Xue Qin

Author 4: Fawang Guo

Keywords: Beach litter detection; QSB-YOLO; YOLOv7; Quantization-Aware RepVGG; a simple; parameter-free attention module; bidirectional feature pyramid network

Paper 36: SECI Model Design with a Combination of Data Mining and Data Science in Transfer of Knowledge of College Graduates’ Competencies

Abstract: One of the methods in knowledge management that can be used is the SECI Model. The SECI Model transfers tacit and explicit knowledge in each quadrant. However, without using tools, the transfer of technical knowledge will experience various obstacles. These obstacles included limited knowledge of the informants, difficulty in translating what was conveyed by the informants, limited time and opportunities, and unclear results obtained. The transfer of knowledge needed by college institutions is in the form of input from graduates who have graduated from college institutions. Graduates' knowledge must be obtained to determine whether their competence is following their respective fields of knowledge. Information technology can help overcome technical problems in transferring knowledge, including the problem of large amounts of data. Data science will deliver results from a combination of technology and mathematics. Meanwhile, data mining, especially with classification, grouping, and association functions, can provide a clear picture of the needs of higher education institutions for the knowledge of their graduates to assess the curriculum that has been provided so far. The design formulation of the SECI model and the implementation of this data mining use an empirical approach through observation and experimentation with quantitative data, as well as theoretical thinking in supporting the development of the model development concept. Data mining and data science will clarify processes in the SECI Model quadrant regarding technological tools in the context of knowledge transfer in a circular manner between tacit and explicit, in order to be more directed and precise. Information extracted from graduate competencies can assist college institutions in formulating future strategies in the academic field, especially the curriculum in study program. This result will impact students in the future, where the developed curriculum will focus more on the results of the input of graduate students.

Author 1: Mardiani

Author 2: Ermatita

Author 3: Samsuryadi

Author 4: Abdiansah

Keywords: Model SECI; data mining; data science; competence of graduates

Paper 37: Enhancing Cloud Security: An Optimization-based Deep Learning Model for Detecting Denial-of-Service Attacks

Abstract: DoS (Denial-of-Service) attacks pose an imminent threat to cloud services and could cause significant financial and intellectual damage to cloud service providers and their customers. DoS attacks can also result in revenue loss and security vulnerabilities due to system disruptions, interrupted services, and data breaches. However, despite machine learning methods being the research subject for detecting DoS attacks, there has not been much advancement in this area. As a consequence of this, there is a requirement for additional research in this field to create the most effective models for the detection of DoS attacks in cloud-based environments. This research paper suggests a deep convolutional generative adversarial network as an optimization-based deep learning model for identifying DoS bouts in the cloud. The proposed model employs Deep Convolutional Generative Adversarial Networks (DCGAN) to comprehend the spatial and temporal features of network traffic data, thereby enabling the attack detection of patterns indicative of DoS assaults. Furthermore, to make the DCGAN more accurate and resistant to attacks, it is trained on a massive collection of network traffic data. Moreover, the model is optimized via backpropagation and stochastic gradient descent to lessen the loss function, quantifying the gap between the simulated and observed traffic volumes. The testing findings prove that the suggested model is superior to the most recent technology methods for identifying cloud-based DoS assaults in Precision and the rate of false positives.

Author 1: Lamia Alhazmi

Keywords: DOS attack; cloud database; generative adversarial networks; attack detection; security threats

Paper 38: Inspection System for Glass Bottle Defect Classification based on Deep Neural Network

Abstract: The problem of defects in glass bottles is a significant issue in glass bottle manufacturing. There are various types of defects that can occur, including cracks, scratches, and blisters. Detecting these defects is crucial for ensuring the quality of glass bottle production. The inspection system must be able to accurately detect and automatically determine that the defects in a bottle affect its appearance and functionality. Defective bottles must be identified and removed from the production line to maintain product quality. This paper proposed glass bottle defect classification using Convolutional Neural Network with Long Short-Term Memory (CNNLSTM) and instant base classification. CNNLSTM is used for feature extraction to create a representation of the class data. The instant base classification predicts anomalies based on the similarity of representations of class data. The convolutional layer of the CNNLSTM method incorporates a transfer learning algorithm, using pre-trained models such as ResNet50, AlexNet, MobileNetV3, and VGG16. In this experiment, the results were compared with ResNet50, AlexNet, MobileNetV3, VGG16, ADA, Image threshold, and Edge detection methods. The experimental results demonstrate the effectiveness of the proposed method, achieving high classification accuracies of 77% on the body dataset, 95% on the neck dataset, and an impressive 98% on the rotating dataset.

Author 1: Niphat Claypo

Author 2: Saichon Jaiyen

Author 3: Anantaporn Hanskunatai

Keywords: Convolution neural network; glass bottle; defect detection; long shot-term memory; inspection machine

Paper 39: Anomalous Taxi Trajectory Detection using Popular Routes in Different Traffic Periods

Abstract: Anomalous trajectory detection is an important approach to detecting taxi fraud behaviors in urban traffic systems. The existing methods usually ignore the integration of the trajectory access location with the time and trajectory structure, which incorrectly detects normal trajectories that bypass the congested road as anomalies and ignores circuitous travel of trajectories. Therefore, this study proposes an anomalous trajectory detection algorithm using the popular routes in different traffic periods to solve this problem. First, to obtain popular routes in different time periods, this study divides the time according to the time distribution of the traffic trajectories. Second, the spatiotemporal frequency values of the nodes are obtained by combining the trajectory point moments and time span to exclude the interference of the temporal anomaly trajectory on the frequency. Finally, a gridded distance measurement method is designed to quantitatively measure the anomaly between the trajectory and the popular routes by combining the trajectory position and trajectory structure. Extensive experiments are conducted on real taxi trajectory datasets; the results show that the proposed method can effectively detect anomalous trajectories. Compared to the baseline algorithms, the proposed algorithm has a shorter running time and a significant improvement in F-Score, with the highest improvement rate of 7.9%, 5.6%, and 10.7%, respectively.

Author 1: Lina Xu

Author 2: Yonglong Luo

Author 3: Qingying Yu

Author 4: Xiao Zhang

Author 5: Wen Zhang

Author 6: Zhonghao Lu

Keywords: Anomalous trajectory detection; time periods; popular routes; gridded distance

Paper 40: The Model of Stroke Rehabilitation Service and User Demand Matching

Abstract: This article focuses on matching stroke rehabilitation services, and patient needs through the interconnection between patient demand and rehabilitation service capabilities. A solution is proposed based on the KJ, fuzzy AHP, and QFD methods to address this problem. Specifically, the KJ method categorizes user needs, and the fuzzy AHP method calculates weights and rankings. Furthermore, rehabilitation service capability indicators are developed, and the QFD method is applied to match customer needs with rehabilitation service capability indicators. The service indicator value is constructed through mapping relationships, and the rehabilitation service capability value is obtained by adding up the results. The best matching scheme is predicted by comparing rehabilitation service capability values of service alternatives. The success of the model has been proven by examining the case. It has helped patients and service organizations find suitable caregivers. The research results illustrate that the proposed model can effectively address the problem of stroke rehabilitation services and patient needs matching and has practical value and potential applications. Therefore, this research is significant in enhancing the quality of stroke rehabilitation services and patient satisfaction and provides a reference value for future studies of similar issues.

Author 1: Hua Wei

Author 2: Ding-Bang Luh

Author 3: Yue Sun

Author 4: Xiao-Hong Mo

Author 5: Yu-Hao Shen

Keywords: Stroke; rehabilitation services; user needs; matching model

Paper 41: Detection of Protective Apparatus for Municipal Engineering Construction Personnel Based on Improved YOLOv5s

Abstract: With the rapid economic development, the government has increased investment in municipal construction, which usually takes a long time, involves many open-air operations, and is affected by cross-construction, traffic, climate and environment, and so on. The safety protection of urban construction workers has been a concern. In this paper, an improved algorithm based on YOLOv5s for the simultaneous detection of helmets and reflective vests is proposed for municipal construction management. First, a new data enhancement method, Mosaic-6, is used to improve the model's ability to learn local features. Second, the SE attention mechanism is introduced in the focus module to expand the perceptual field, strengthen the degree of association between channel information and the detection target, and improve the detection accuracy. Finally, the features of small-scale targets are interacted and fused in multiple dimensions according to the Swin transformer network structure. The experimental results show that the improved algorithm achieves accuracy, recall, and mean accuracy rates of 98.5%, 97.0%, and 92.7%, respectively. These results show an average improvement of 3.4 percentage points in mean accuracy compared to the basic YOLOv5s. This study provides valuable insights for further research in the area of urban engineering security and protection.

Author 1: Shuangyuan Li

Author 2: Yanchang Lv

Author 3: Mengfan Li

Author 4: Zhengwei Wang

Keywords: YOLOv5s; hard hat; reflective vest; simultaneous detection

Paper 42: A Review of Fake News Detection Models: Highlighting the Factors Affecting Model Performance and the Prominent Techniques Used

Abstract: In recent times, social media has become the primary way people get news about what is happening in the world. Fake news surfaces on social media every day. Fake news on social media has harmed several domains, including politics, the economy, and health. Additionally, it has negatively affected society's stability. There are still certain limitations and challenges even though numerous studies have offered useful models for identifying fake news in social networks using many techniques. Moreover, the accuracy of detection models is still notably poor given we deal with a critical topic. Despite many review articles, most previously concentrated on certain and repeated sections of fake news detection models. For instance, the majority of reviews in this discipline only mentioned datasets or categorized them according to labels, content, and domain. Since the majority of detection models are built using a supervised learning method, it has not been investigated how the limitations of these datasets affect detection models. This review article highlights the most significant components of the fake news detection model and the main challenges it faces. Data augmentation, feature extraction, and data fusion are some of the approaches explored in this review to improve detection accuracy. Moreover, it discusses the most prominent techniques used in detection models and their main advantages and disadvantages. This review aims to help other researchers improve fake news detection models.

Author 1: Suhaib Kh. Hamed

Author 2: Mohd Juzaiddin Ab Aziz

Author 3: Mohd Ridzwan Yaakub

Keywords: Fake news detection; social media; data augmentation; feature extraction; multimodal fusion

Paper 43: A Dynamic Intrusion Detection System Capable of Detecting Unknown Attacks

Abstract: In recent years, deep learning-based network intrusion detection systems (IDS) have shown impressive results in detecting attacks. However, most existing IDS can only recognize known attacks that were included in their training data. When faced with unknown attacks, these systems are often unable to take appropriate actions and incorrectly classify them into known categories, leading to reduced detection performance. Furthermore, as the number and types of network attacks continue to increase, it becomes challenging for these IDS to update their model parameters promptly and adapt to new attack scenarios. To address these issues, this paper introduces a dynamic intrusion detection system, Dynamic Unknown Attack Intrusion Detection System (DUA-IDS). This system aims to learn and detect unknown attacks effectively. DUA-IDS comprises three components: Feature Extractor: This component employs CNN and Transformer models to extract data features from various perspectives. Threshold-Based Classifier: The second part utilizes the nearest mean rule of samples to classify known and unknown attacks, enabling the distinction between them. Dynamic Learning Module: The third part incorporates data playback and knowledge distillation techniques to retain existing category knowledge while continuously learning new attack categories. To assess the effectiveness of DUA-IDS, this paper conducted experiments using the UNSW-NB15 public dataset. The experimental results show that DUA-IDS improves the classification accuracy of flow network data with unknown traffic attacks. Can accurately distinguish unknown traffic and correctly classify known traffic. When dynamically learning unknown traffic, the classification accuracy of previously learned known traffic is less affected. This indicates the advantages of DUA-IDS in detecting unknown attacks and learning new attack categories.

Author 1: Na Xing

Author 2: Shuai Zhao

Author 3: Yuehai Wang

Author 4: Keqing Ning

Author 5: Xiufeng Liu

Keywords: Intrusion detection systems; transformer; DUA-IDS; data playback; variable perspectives features; Knowledge distillation

Paper 44: Computational Framework for Analytical Operation in Intelligent Transportation System using Big Data

Abstract: Intelligent Transportation System (ITS) is the future of the current transport scheme. It is meant to incorporate an intelligent traffic management operation to offer vehicles more safety and valuable traffic-related information. A review of existing approaches showcases the implementation of varied scattered schemes where analytical operation is mainly emphasized. However, some significant shortcomings are witnessed in efficiently managing complex traffic data. Therefore, the proposed system introduces a novel computational framework with a joint operation toward analytical processing using big data targeting to manage raw and complex traffic data efficiently. As a novel feature, the model introduces a data manager who can manage the complex traffic stream, followed by decentralized traffic management, that can identify and eliminate artefacts using statistical correlation. Finally, predictive modelling is incorporated to offer knowledge discovery with the highest accuracy. The simulation outcome shows that Random Forest excels with 99% accuracy, which is the highest of all existing machine learning approaches, along with the accomplishment of 11.77% reduced overhead, 1.3% of reduced delay, and 67.47% reduced processing time compared to existing machine learning approaches.

Author 1: Mahendra G

Author 2: Roopashree H. R

Keywords: Intelligent transportation system; traffic managemen; machine learning; artifacts; prediction

Paper 45: Comparison of Multi-layer Perceptron and Support Vector Machine Methods on Rainfall Data with Optimal Parameter Tuning

Abstract: This study describes the search for optimal hyperparameter values in rainfall data in 49 cities in Australia, consisting of 145,460 records with 22 features. The process eliminates missed values and selects 16 numeric type features as input features and one feature (Rain Tomorrow) as output feature. It is processed using the Multi-Layer Perceptron (MLP) and Support Vector Machine (SVM) methods based on Three Best Accuration (3BestAcc) and Best Three Nearest Neighbors (3BestNN). The results showed that the SVM kernel linear method gave an average accuracy value of 0.85586 and was better than the MLP method with an accuracy of 0.854.

Author 1: Marji

Author 2: Agus Widodo

Author 3: Marjono

Author 4: Wayan Firdaus Mahmudy

Author 5: Maulana Muhamad Arifin

Keywords: Rainfall; MLP; SVM; optimal

Paper 46: A Vehicle Classification System for Intelligent Transport System using Machine Learning in Constrained Environment

Abstract: Vehicle type classification has an extensive variety of applications which include intelligent parking systems, traffic flow-statistics, toll collecting system, vehicle access control, congestion management, security system and many more. These applications are designed for reliable and secure transportation. Vehicle classification is one of the major challenges of these applications particularly in a constrained environment. The constrained environment in the real world put a limit on data quality due to noise, poor lightning condition, low resolution images and bad weather conditions. In this research, we build a more practical and more robust vehicle type classification system for real world constrained environment with promising results and got a validation accuracy of 90.85 and a testing accuracy of 87%. To this end, we design a framework of vehicle type classification from vehicle images by using machine learning. We investigate the deep learning method Convolutional neural network (CNN), a specific type of neural networks. CNN is biologically inspired with multi-layer feed forward neural networks. It can learn automatically at several stages of invariant features for the particular chore. For evaluation, we also compared the performance of our model with the performance of other machine learning algorithms like Naïve Bayes, SVM and Decision Trees.

Author 1: Ahmed S. Alghamdi

Author 2: Talha Imran

Author 3: Khalid T. Mursi

Author 4: Atika Ejaz

Author 5: Muhammad Kamran

Author 6: Abdullah Alamri

Keywords: Vehicle classification; intelligent transport system; deep learning; machine learning; CNN; digital image processing

Paper 47: Drug Resistant Prediction Based on Plasmodium Falciparum DNA-Barcoding using Bidirectional Long Short Term Memory Method

Abstract: Malaria disease mostly affects children and causes death every year. Multiple factors of the disease due to failure in treatment, including anti-malaria drug resistance. The resistance is caused by a decrease in the efficacy of the drug against Plasmodium parasites. Therefore, we proposed a computational approach using deep learning methods to predict anti-malarial drug resistance based on genetic variants of the Plasmodium falciparum through DNA barcoding. The DNA Barcode, organism identification from Plasmodium, is employed as data set for predicting the anti-malaria drug resistance. As a univariate amino acid sequence, it is transformed to numerical value data for building classifier model. It is constructed into a classifier model for prediction using Bidirectional Long Term-Short Memory (Bi-LSTM). This algorithm is extended from LSTM by two directions. In the first stage, the sequence is encoded into numerical data as input data for the method using sigmoid activation loss function. Then binary cross entropy is addressed to define the class, resistance or sensitivity. The final stage is applied by tuning hyper-parameter using Adaptive Moment Estimation optimizer to get the best performance. The experimental results show that Bi-LSTM as the proposed method achieves high performance for resistance prediction including precision, recall, and f1-score.

Author 1: Lailil Muflikhah

Author 2: Nashi Widodo

Author 3: Novanto Yudistira

Author 4: Achmad Ridok

Keywords: Drug resistant; plasmodium falciparum; Bi-LSTM; deep learning

Paper 48: Toward Modeling Trust Cyber-Physical Systems: A Model-based System Engineering Method

Abstract: Developing trust in cyber-physical systems (CPSs) is a challenging task. Trust in CPS is needed for carrying out their intended duties and is reasonably safe from misuse and intrusion; it also enforces the applicable security policy. As an example, medical smart devices, many researches have found that trust is a key factor in explaining the relationship between individual beliefs about technological attributes and their acceptance behavior; and have associated medical device failures with severe patient injuries and deaths. The cyber-physical system is considered a trust system if the principles of security and safety, confidentiality, integrity, availability, and other attributes are assured. However, a lack of sufficient analysis of such systems, as well as appropriate explanation of relevant trust assumptions, may result in systems that fail to completely realize their functionality. The existing research does not provide suitable guidance for a systematic procedure or modeling language to support such trust-based analysis. The most pressing difficulties are achieving trust by design in CPS and systematically incorporating trust engineering into system development from the start of the system life cycle. Still, there is a need for a strategy or standard model to aid in the creation of a safe, secure, and trustworthy CPS. Model-based system engineering (MBSE) approaches for trust cyber-physical systems are a means to address system trustworthiness design challenges. This work proposes a practical and efficient MBSE method for constructing trust CPS, which provides guidance for the process of trustworthiness analysis. The SysML-based profile is supplied, together with recommendations on which approach is required at each process phase. The MBSE method is proven by expanding the autonomous car SysML and UML diagrams, and we show how trust considerations are integrated into the system development life cycle.

Author 1: Zina Oudina

Author 2: Makhlouf Derdour

Keywords: Cyber Physical Systems (CPSs); trust CPS; system engineering (SE); model-based system engineering (MBSE); SysML

Paper 49: Construction of an Ontology-based Document Collection for the IT Job Offer in Morocco

Abstract: Information Technology (IT) job offers are available on the web in a heterogeneous way. It is difficult for a candidate looking for an IT job to retrieve the exact information they need to locate the ideal match for their profile, without wasting time on useless searches. Traditional IT job search systems are based on simple keywords that are generally not adapted to provide detailed answers because they do not take into account semantic links. In this article, an ontology is developed to meet the expectations of IT profiles from the IT job descriptions accumulated and pre-annotated using the UBIAI tool. The classes and subclasses of the ontology are designed using the Protégé 5.5.0 editor. Then the properties of objects and data are defined to improve the ontology. The ontology results are validated using DL queries by asking a number of questions to retrieve the requested information for each IT profile, and the ontology answers all these questions adequately. Finally, Various plugins are used to display an ontology in a graphical representation.

Author 1: Zineb Elkaimbillah

Author 2: Bouchra El Asri

Author 3: Mounia Mikram

Author 4: Maryem Rhanoui

Keywords: Ontology; IT job descriptions; semantic links; DL query; protégé 5.5.0

Paper 50: Enhancing User Experience Via Calibration Minimization using ML Techniques

Abstract: Electromyogram (EMG) signals are used to recognize gestures that could be used for prosthetic-based and hands-free human computer interaction. Minimizing calibration times for users while preserving the accuracy, is one of the main challenges facing the practicality, user acceptance and spread of upper limb movements’ detection systems. This paper studies the effect of minimized user involvement, thus user calibration time and effort, on the user-independent system accuracy. It exploits time based features extracted from EMG signals. One-versus-all kernel based Support Vector Machine (SVM) and K Nearest Neighbors (KNN) are used for classification. The experiments are conducted using a dataset having five subjects performing six distinct movements. Two experiments performed, one with complete user dependence condition and the other with the partial dependence. The results show that the involvement of at least two samples, representing around 2% of sample space, increase the performance by 62.6% in case of SVM, achieving accuracy result equal to 89.6% on average; while the involvement of at least three samples, representing around 3% of sample space, cause the increase by 50.6% in case of KNN, achieving accuracy result equal to 78.2% on average. The results confirmed the great impact on system accuracy when involving only small number of user samples in the model-building process using traditional classification methods.

Author 1: Sarah N. AbdulKader

Author 2: Taha M. Mohamed

Keywords: EMG signals; user independence; EMG user acceptance; HCI; movement classification; calibration minimization

Paper 51: A New Approach Method for Multi Classification of Lung Diseases using X-Ray Images

Abstract: Lung disease is one of the most common diseases in today's society. This lung disease's treatment is frequently postponed. This is usually due to a lack of understanding about proper treatment and a lack of clear information about lung disease. Reading the correct X-ray images, which is usually done by experts who are familiar with these X-rays, is one method of detecting lung disease. However, the results of this diagnosis are dependent on the expert's practice schedule and take a long time. This study aims to classify lung disease images using preprocessing, augmentation, and multimachine learning methods, with the goal of achieving high classification performance accuracy with multi-class lung disease. The classification ExtraTrees was obtained from experimental results with unbalanced datasets using a balancing process with augmentation. Precision, Recall, Fi-Score, and Accuracy are 100% for training and testing data 89% for Precision, 88% for Recall, 87 for Fi-Score, and 85% for Accuracy outperform other machine learning models such as Kneighbors, Support Vector Machine (SVM), and Random Forest in classifying lung diseases. The conclusion from this research is that the machine learning approach can detect several lung diseases using X-ray images.

Author 1: Sri Heranurweni

Author 2: Andi Kuniawan Nugroho

Author 3: Budiani Destyningtias

Keywords: Augmentation; machine learning; lung disease; prepossessing

Paper 52: Optimizing YOLO Performance for Traffic Light Detection and End-to-End Steering Control for Autonomous Vehicles in Gazebo-ROS2

Abstract: Autonomous driving has become a popular area of research in recent years, with accurate perception and recognition of the environment being critical for successful implementation. Traditional methods for recognizing and controlling steering rely on the color and shape of traffic lights and road lanes, which can limit their ability to handle complex scenarios and variations in data. This paper presents an optimization of the You Only Look Once (YOLO) object detection algorithm for traffic light detection and end-to-end steering control for lane-keeping in the simulation environment. The study compares the performance of YOLOv5, YOLOv6, YOLOv7, and YOLOv8 models for traffic light signal detection, with YOLOv8 achieving the best results with a mean Average Precision (mAP) of 98.5%. Additionally, the study proposes an end-to-end convolutional neural network (CNN) based steering angle controller that combines data from a classical proportional integral derivative (PID) controller and the steering angle controller from human perception. This controller predicts the steering angle accurately, outperforming conventional open-source computer vision (OpenCV) methods. The proposed algorithms are validated on an autonomous vehicle model in a simulated Gazebo environment of Robot Operating System 2 (ROS2).

Author 1: Hoang Tran Ngoc

Author 2: Khang Hoang Nguyen

Author 3: Huy Khanh Hua

Author 4: Huynh Vu Nhu Nguyen

Author 5: Luyl-Da Quach

Keywords: Yolo models; PID; CNN; gazebo; ROS2; traffic-light; lane-keeping; autonomous

Paper 53: DefBDet: An Intelligent Default Borrowers Detection Model

Abstract: The growing popularity and availability of online lending platforms have attracted more borrowers and lenders. There have been several studies focusing on analyzing loan risks in the financial industry, however, defaulting loans still remains an issue that needs more attention. Hence, this research aims to develop an intelligent prediction model that is able to predict risky loans and default borrowers, named the Default Borrowers Detection Model (DefBDet). We seek to help loan lending platforms to approve lending loans to those who are expected to comply with re-payments at the agreed time. Previous works developed a binary classification prediction model (either default or repaid loan). Repaid loans include loans being repaid on or after the loan deadline date. DefBDet, on the other hand, is a novel model, it can predict a loan status based on a multi-classification bases rather than a binary class bases. Hence, it can additionally identify expected late repaid loans, so that special conditions are assigned before loan being approved. This study employs seven different Machine Learning models, using a real-world dataset from 2009-2022 consisting of around 255k loan requests. Statistical measures such as Recall, Precision, and F-measure have been used for models' evaluation. Results show that Random Forest has achieved the highest performance of 85%.

Author 1: Fooz Alghamdi

Author 2: Nora Alkhamees

Keywords: Default borrowers; default loans; loan risks; machine learning models; prediction model

Paper 54: Effects of Training Data on Prediction Model for Students' Academic Progress

Abstract: The ability to predict students’ academic performance before the start of the class with credible accuracy could significantly aid the preparation of effective teaching and learning strategies. Several studies have been conducted to enhance the performance of prediction models by emphasizing three key factors: developing effective prediction algorithms, identifying significant predictor variables, and developing preprocessing techniques. Importantly, none of these studies focused on the effect of using different types of training data on the performance of prediction models. Therefore, this study was conducted to evaluate the effects of differences in training data on the performance of a prediction model designed to monitor students’ academic progress. The findings showed that the performance of the prediction model was strongly influenced by the heterogeneity of the values of the predictor variables, which should accommodate all the existing possibilities. It was also discovered that the application of training data with different characteristics and sizes did not improve the performance of the prediction model when its heterogeneity was not representative.

Author 1: Susana Limanto

Author 2: Joko Lianto Buliali

Author 3: Ahmad Saikhu

Keywords: Decision tree; effects of training data; heterogeneity; prediction; students’ academic performance

Paper 55: A Novel Framework for Detecting Network Intrusions Based on Machine Learning Methods

Abstract: In the rapidly evolving landscape of cyber threats, the efficacy of traditional rule-based network intrusion detection systems has become increasingly questionable. This paper introduces a novel framework for identifying network intrusions, leveraging the power of advanced machine learning techniques. The proposed methodology steps away from the rigidity of conventional systems, bringing a flexible, adaptive, and intuitive approach to the forefront of network security. This study employs a diverse blend of machine learning models including but not limited to, Convolutional Neural Networks (CNNs), Support Vector Machines (SVMs), and Random Forests. This research explores an innovative feature extraction and selection technique that enables the model to focus on high-priority potential threats, minimizing noise and improving detection accuracy. The framework's performance has been rigorously evaluated through a series of experiments on benchmark datasets. The results consistently surpass traditional methods, demonstrating a remarkable increase in detection rates and a significant reduction in false positives. Further, the machine learning-based model demonstrated its ability to adapt to new threat landscapes, indicating its suitability in real-world scenarios. By marrying the agility of machine learning with the concreteness of network intrusion detection, this research opens up new avenues for dynamic and resilient cybersecurity. The framework offers an innovative solution that can identify, learn, and adapt to evolving network intrusions, shaping the future of cyber defense strategies.

Author 1: Batyrkhan Omarov

Author 2: Nazgul Abdinurova

Author 3: Zhamshidbek Abdulkhamidov

Keywords: Attack detection; intrusion detection; machine learning; information security; artificial intelligence

Paper 56: A Dynamic Model for Risk Assessment of Cross-Border Fresh Agricultural Supply Chain

Abstract: The cross-border trade of Fresh Agricultural Products (FAP) is widespread in the current society, and the demand for it is also increasing. The cross-border fresh agricultural product Supply Chain (SP) itself has strong complexity and high costs, and it also bears many risks. In order to alleviate the adverse impact of risk factors interfering with cross-border fresh agricultural product SPs and improve the overall SP efficiency, this study proposes a system dynamics model based on cross-border fresh agricultural product risk factors. The experiment first studied the possible risk factors in the SP of FAP. After discussing the causal relationship between possible risks, subjective and objective weighting methods were introduced to weight risk factors. After that, a system dynamics model of the cross-border fresh agricultural product SP was constructed for the purpose of enhancing product quality and the overall efficiency of the SP. In the system dynamics model constructed, risk factors are introduced for simulation experiments. It is demonstrated that the suggested model can truly reflect the dynamic changes of the actual SP, and can obtain the operational rules of the system.

Author 1: Honghong Zhai

Keywords: Cross-border fresh agricultural products; supply chain management; risk identification; system dynamics model; risk weighting

Paper 57: Design of an Educational Platform for Professional Development of Teachers with Elements of Blockchain Technology

Abstract: This paper presents an in-depth examination of the development and implementation of an innovative platform for teacher professional development, incorporating features of blockchain technology. The platform manifests a revolutionary step in enhancing teacher training, creating a secure, transparent, and decentralized approach for maintaining continuous professional development records. Using blockchain's inherent properties, the platform ensures immutable record-keeping and instills credibility in teachers' career progression, empowering educators through direct ownership of their professional development milestones. Additionally, the platform fosters a culture of lifelong learning, encouraging educators to actively engage in their professional growth, while providing reliable evidence of their achievements. Alongside highlighting the design aspects of the platform, the paper delves into potential challenges and solutions associated with the incorporation of blockchain technology into educational contexts. Through this innovative intersection of technology and education, the platform showcases the potential of blockchain in reshaping and enriching professional development strategies for teachers, thereby elevating educational standards and practices across the board.

Author 1: Aivar Sakhipov

Author 2: Talgat Baidildinov

Author 3: Madina Yermaganbetova

Author 4: Nurzhan Ualiyev

Keywords: Blockchain; professional development; artificial intelligence; teaching; learning

Paper 58: Predicting Maintenance Labor Productivity in Electricity Industry using Machine Learning: A Case Study and Evaluation

Abstract: Predicting maintenance labor productivity is crucial for effective planning and decision-making in the electricity industry. This paper aims at predicting maintenance labor productivity using various machine learning methods, utilizing a real-world case study from the electricity industry. Additionally, the study evaluates the performance of the employed machine learning methods. To meet this objective, 1750 productivity measures have been used to train (80%) and test (20%) prediction models using Artificial Neural Networks, Support Vector Machines, Random Forest, and Multiple Linear Regression methods. The models' performance was evaluated based on the mean squared error, mean absolute percentage error, and testing time. The results indicated that the Artificial Neural Networks model - specifically, a feedforward network with a backpropagation algorithm - outperformed the other models (Multiple Linear Regression, Support Vector Machines, Random Forest). These results highlight the effectiveness of machine learning, particularly the Artificial Neural Networks prediction model, as an invaluable tool for decision-makers in the electricity industry, aiding in more effective maintenance planning and potential productivity improvement.

Author 1: Mariam Alzeraif

Author 2: Ali Cheaitou

Author 3: Ali Bou Nassif

Keywords: Productivity; machine learning; maintenance; prediction; ANN

Paper 59: Web Phishing Classification Model using Artificial Neural Network and Deep Learning Neural Network

Abstract: Phishing is an online crime in which a cybercriminal tries to persuade internet users to reveal important and sensitive personal information, such as bank account details, usernames, passwords, and social security numbers, to the phisher, usually for mean purposes. The target victim of the fraud suffers a financial loss, as well as the loss of personal information and reputation. Therefore, it is essential to identify an effective approach for phishing website classification. Machine learning approaches have been applied in the classification of phishing websites in recent years. The objectives of this research are to classify phishing websites using artificial neural network (ANN) and convolutional neural network (CNN) and then compare the results of the models. This study uses a phishing website dataset collected from the machine learning database, University of California, Irvine (UCI). There were nine input attributes and three output classes that represent types of websites either legitimate, suspicious, or phishing. The data was split into 70% and 30% for training and testing purposes, respectively. The results indicate that the modified ANN with Rectified Linear Unit (ReLU) activation function model outperforms other models by achieving the least average of root mean square error (RMSE) value for testing which is 0.2703, while the CNN model produced the least average RMSE for training which is 0.2631. ANN with Sigmoid activation function model obtained the highest average RMSE of 0.3516 for training and 0.3585 for testing.

Author 1: Noor Hazirah Hassan

Author 2: Abdul Sahli Fakharudin

Keywords: Phishing website; classification; artificial neural network; convolutional neural network; machine learning

Paper 60: Purchase Intention and Sentiment Analysis on Twitter Related to Social Commerce

Abstract: Social commerce is a digital and efficient solution to transform existing commerce and address contemporary issues. TikTok Shop, a popular and trending social commerce platform, competes with established competitors like Facebook Marketplace and Instagram Shop. TikTok Shop offers benefits and incentives to attract users for both sales and product purchases. In this study, various algorithmic approaches such as Naïve Bayes, K-Nearest Neighbor, Support Vector Machine, Logistic Regression, Decision Tree, Random Forest, LGBM Boost, Ada Boost, and Voting Classifier are utilized to analyze and compare sentiments expressed on Twitter regarding Facebook, Instagram, and TikTok. The aim is to determine the methods with the best performance and identify the social commerce platform with the highest purchase intention and positive sentiment. The results indicate that TikTok has more positive sentiment than Facebook and Instagram at 93.07% with the best-performing classification model, Decision Tree. In conclusion, TikTok exhibits the highest positive sentiment percentage, indicating a greater number of positive reviews compared to Facebook and Instagram. According to the theory of evaluation scores for measuring model performance, values above 0.90 represent models with good performance.

Author 1: Muhammad Alviazra Virgananda

Author 2: Indra Budi

Author 3: Kamrozi

Author 4: Ryan Randy Suryono

Keywords: Algorithm; machine learning; sentiment; social commerce

Paper 61: Ensemble Deep Learning (EDL) for Cyber-bullying on Social Media

Abstract: Cyber-bullying is a growing problem in the digital age, affecting millions of people worldwide. Deep learning algorithms have the potential to assist in identifying and combating Cyber-bullying by detecting and classifying harmful messages. This paper uses two Ensemble deep learning (EDL) models to detect Cyber-bullying on text data, images and videos—and an overview of Cyber-bullying and its harmful effects on individuals and society. The advantages of using deep learning algorithms in the fight against Cyber-bullying include their ability to process large amounts of data and learn and adapt to new patterns of Cyber-bullying behaviour. For text data, firstly, a pre-trained model BERT (Bidirectional Encoder Representations from Transformers) is used to train on cyber-bullying text data. The next step describes the data pre-processing and feature extraction techniques required to prepare data for deep learning algorithms. We also discuss the different types of deep learning algorithms that can be used for Cyber-bullying detection, including convolutional neural networks (CNNs), recurrent neural networks (RNNs), and deep belief networks (DBNs). This paper combines the sentiment analysis model, such as Aspect-based Sentiment Analysis (ABSA), for classifying bullying messages. Deep Neural network (DNN) used the classification of Cyber-bullying images and videos. Experiments were conducted on three datasets such as Twitter (Kaggle), Images (Online), and Videos (Online). Datasets are collected from various online sources. The results demonstrate the effectiveness of EDL and DNN in detecting Cyber-bullying in terms of detecting bullying data from relevant datasets. The EDL and DNN obtained an accuracy of 0.987, precision of 0.976, F1-score of 0.975, and recall of 0.971 for the Twitter dataset. The performance of Ensemble CNN brought an accuracy of 0.887, precision of 0.88, F1-score of 0.88, and recall of 0.887 for the Image dataset. For the video dataset, the performance of Ensemble CNN is an accuracy of 0.807, precision of 0.81, F1-score of 0.82, and recall of 0.81. Future research should focus on developing more accurate and efficient deep learning algorithms for Cyber-bullying detection and investigating the ethical implications of using such algorithms in practice.

Author 1: Zarapala Sunitha Bai

Author 2: Sreelatha Malempati

Keywords: Cyber bullying; ensemble deep learning (EDL); convolutional neural networks (CNNs); recurrent neural networks (RNNs); deep belief networks (DBNs)

Paper 62: A Novel Software Quality Characteristic Recommendation Model to Handle the Dynamic Requirements of Software Projects that Improves Service Quality and Cost

Abstract: The software is created and constructed to address particular issues in the applied field. In this context, there is a need to be aware of the crucial characteristics to assess the quality of software. But not all software requires checking all the quality-of-service parameters, resulting in effort loss and time consumption. Therefore, it is required to develop software quality characteristics recommendation model to address and resolve the issue. The proposed work involved in this paper can be subdivided into three main parts (1) a review of popular software quality models and their comparison to create a complete set of predictable, and (2) the design of an ML-based recommendation model for recommending the software quality model and software quality characteristics (3) performance analysis. The proposed recommendation system utilizes the different software quality of service attributes as well as the software attributes where these models are suitably applied to satisfy the demands. Profiling of applications and their essential requirements have been performed Based on the different quality of service parameters and the requirements of applications. These profiles are learned by machine learning algorithms for distinguishing the application-based requirement and recommending the essential attributes. The implementation of the proposed technique has been done using Python technology. The simulation aims to demonstrate how to minimize the cost of software testing and improve time and accuracy by utilizing the appropriate quality matrix. Finally, a conclusion has been drawn and the future extension of the proposed model has been reported.

Author 1: Kamal Borana

Author 2: Meena Sharma

Author 3: Deepak Abhyankar

Keywords: Recommendation system; software quality model; ML (Machine Learning); quality matrix; software quality characteristics

Paper 63: Enhancing Facemask Detection using Deep learning Models

Abstract: Face detection and mask detection are critical tasks in the context of public safety and compliance with mask-wearing protocols. Hence, it is important to track down whoever violated rules and regulations. Therefore, this paper aims to implement four deep learning models for face detection and face with mask detection: MobileNet, ResNet50, Inceptionv3, and VGG19. The models are evaluated based on precision and recall metrics for both face detection and face with mask detection tasks. The results indicate that the proposed model based on ResNet50 achieves superior performance in face detection, demonstrating high precision (99.4%) and recall (98.6%) values. Additionally, the proposed model shows commendable accuracy in mask detection. MobileNet and Inceptionv3 provide satisfactory results, while the proposed model based on VGG19 excels in face detection but shows slightly lower performance in mask detection. The findings contribute to the development of effective face mask detection systems, with implications for public safety.

Author 1: Abdullahi Ahmed Abdirahman

Author 2: Abdirahman Osman Hashi

Author 3: Ubaid Mohamed Dahir

Author 4: Mohamed Abdirahman Elmi

Author 5: Octavio Ernest Romo Rodriguez

Keywords: Object detection; deep learning; detection; face detection; mask detection; convolutional neural network

Paper 64: Sequential Model-based Optimization Approach Deep Learning Model for Classification of Multi-class Traffic Sign Images

Abstract: Autonomous vehicles are currently gaining popularity in the future mobility ecosystem. The development of autonomous driving systems is still challenging in the research area of image processing and signal processing. Extensive research work was conducted on various traffic sign datasets. It achieved respectable results, but a robust network structure is still needed to develop to improve the traffic sign recognition (TSR) system. In this research work, there is an alternative approach to designing deep learning models, which are implemented in TSR systems. The proposed deep learning model was also tested with different datasets to obtain the generalized model. The proposed model was based on a convolutional neural network (CNN), and Bayesian Optimization optimizes the model’s hyperparameters to find the best hyperparameters grid. After that, the optimized CNN model was used to classify the traffic sign images from three different datasets, including the German traffic sign recognition benchmark (GTSRB), the Belgium traffic sign classification (BTSC) dataset, and the Chinese traffic sign database, achieving the average accuracy scores of 99.57%, 99.15%, and 99.35%, respectively.

Author 1: Si Thu Aung

Author 2: Jartuwat Rajruangrabin

Author 3: Ekkarut Viyanit

Keywords: Autonomous driving; convolutional neural network; deep learning; traffic sign; optimization

Paper 65: Whale Optimization-Driven Generative Convolutional Neural Network Framework for Anaemia Detection from Blood Smear Images

Abstract: Anaemia is a frequent blood disorder marked by a reduction in the quantity of haemoglobin or the number of red blood cells in the blood. Quick and accurate anaemia detection is crucial for fast action and effective treatment. In this research, we provide a new structure called Whale Optimization-Driven Generative Convolutional Neural Network (WO-GCNN) for the detection of anaemia using blood smear pictures. To increase anaemia detection accuracy, the WO-GCNN system combines the strength of generative models and convolutional neural networks (CNNs). In order to create artificial blood smear images and learn the underlying data distribution, generative models, such as Generative Adversarial Networks (GANs), are used. Improve the functionality of the WO-GCNN system by applying the Whale Optimisation Algorithm (WOA), which is based on the hunting behaviours of humpback whales. To create the optimal set of CNN weights, the WOA effectively achieves a compromise between exploitation and exploration. The WO-GCNN framework accelerates convergence speed and increases overall performance of anaemia detection by incorporating the WOA into the training process. On a sizable dataset of blood smear pictures obtained from clinical settings, we assess the suggested WO-GCNN system. A highly accurate and effective approach for the early identification of anaemia is produced by combining generative models and CNNs with the WOA optimisation. By enabling early anaemia identification, the proposed WO-GCNN framework has the potential to have a substantial impact on the field of medical image analysis and enhance patient care. It can be a useful tool for medical personnel, supporting them in making decisions and giving anaemia patients urgent interventions.

Author 1: S. Yazhinian

Author 2: Vuda Sreenivasa Rao

Author 3: J. C. Sekhar

Author 4: Suganthi Duraisamy

Author 5: E. Thenmozhi

Keywords: Generative adversarial network; blood smear images; convolutional neural network; anaemia; Whale Optimization

Paper 66: A Transformer-CNN Hybrid Model for Cognitive Behavioral Therapy in Psychological Assessment and Intervention for Enhanced Diagnostic Accuracy and Treatment Efficiency

Abstract: The use of Cognitive Behavioral Therapy (CBT) as a method for psychological assessment and intervention has shown to be quite successful. However, by utilizing advancements in artificial intelligence and natural language processing techniques, the diagnostic precision and therapeutic efficacy of CBT can be significantly improved. For CBT in psychological evaluation and intervention, we suggest a unique Transformer-CNN hybrid model in this work. The hybrid model combines the strengths of the Transformer and Convolutional Neural Network (CNN) architectures. While the CNN model successfully extracts local and global features from the input sequences, the Transformer model accurately captures the contextual dependencies and semantic linkages in the text data. It intends to enhance the model's comprehension and interpretation of the complex linguistic patterns involved in psychological evaluation and intervention by merging these two algorithms. On a sizable collection of clinical text data, which includes patient narratives, treatment transcripts, and diagnostic reports, we undertake comprehensive experiments. The proposed Trans-CNN hybrid model outperformed all other methods with an impressive accuracy of 97%. In diagnosing psychiatric problems, the model shows improved diagnosis accuracy and offers more effective therapy advice. Our hybrid model's automatic real-time monitoring and feedback capabilities also make it possible for prompt intervention and customized care during therapy sessions. By giving doctors a formidable tool for precise evaluation and efficient intervention, the suggested approach has the potential to revolutionize the field of CBT and enhance patient outcomes for mental health. In order to improve the diagnostic precision and therapeutic efficacy of CBT in psychological evaluation and intervention, this work provides a transformational strategy that combines the advantages of the Transformer and CNN architectures.

Author 1: Veera Ankalu Vuyyuru

Author 2: G Vamsi Krishna

Author 3: S. Suma Christal Mary

Author 4: S. Kayalvili

Author 5: Abraheem Mohammed Sulayman Alsubayhay

Keywords: CBT; psychological assessment; intervention; diagnostic accuracy; treatment efficiency; Transformer; CNN; NLP

Paper 67: A Stacking-based Ensemble Framework for Automatic Depression Detection using Audio Signals

Abstract: Mental illnesses are severe obstacles for the global welfare. Depression is a psychological disorder which causes problems to the individual and also to his/her dependents. Machine learning based methods using audio signals can differentiate patterns between healthy and depressive subjects. These methods can assist health care professionals to detect the depression. Literature in depression detection, based on audio signals, used only single classifier, lacks to take advantage of diverse classifiers. The current work combines predictive capabilities of diverse classifiers using stacking method to detect depression. Audio clips are reordered while a predefined paragraph is being read out, for acoustic analysis of speech. The dataset is created which has extended Geneva Minimalistic Acoustic Parameter Set (eGeMAPS) features, that are extracted using openSMILE toolkit. The normalized feature vectors are given as input to multiple classifiers to give an intermediate prediction. These predictions are combined using a meta classifier to form a final outcome. K-Nearest Neighbours (KNN), Naïve Bayes (NB), Support Vector Machine (SVM), and Decision Trees (DT) classifiers are utilised on the normalized feature vector for intermediate predictions and Logistic Regression (LR) is used as meta classifier to predict final outcome. Our proposed method of using diverse classifiers achieved significant accuracy of 79.1%, precision of 83.3%, recall of 76.9% and F1-score of 80% on our dataset. Results are discussed while using stacking method on our dataset, then compared with various baseline methods also while applying on a publicly available bench marking dataset. Our results showed that combining predictive capability of multiple diverse classifiers helps in depression detection.

Author 1: Suresh Mamidisetti

Author 2: A. Mallikarjuna Reddy

Keywords: Health care; depression detection; acoustic features; speech elicitation; feature selection; openSMILE; ensemble methods

Paper 68: Optimizing Port Operations: Synchronization, Collision Avoidance, and Efficient Loading and Unloading Processes

Abstract: This study focuses on optimizing the loading and unloading processes in a port environment by employing synchronization techniques and collision avoidance mechanisms. The objective function of this research aims to minimize the time required for these tasks while ensuring efficient coordination and safety. The obtained results are compared with previous studies, demonstrating significant improvements in overall performance. The synchronized handling systems, including gantries and cranes, along with speed control measures, facilitate streamlined operations, reduced delays, and enhanced productivity. By integrating these strategies, the port achieves better results in terms of task completion time compared to previous methodologies, thereby validating the effectiveness of the proposed approach.

Author 1: Sakhi Fatima Ezzahra

Author 2: Bellat Abdelouahad

Author 3: Mansouri Khalifa

Author 4: Qbadou Mohammed

Keywords: Optimizing; synchronization; collision; efficient; time

Paper 69: Optimized Ensemble of Hybrid RNN-GAN Models for Accurate and Automated Lung Tumour Detection from CT Images

Abstract: The early diagnosis and treatment of lung tumour, the primary cause of cancer-related deaths globally, depend critically on the identification of lung tumours. In this approach, a new method is suggested for detecting lung tumours that combines a Gaussian filter with a hybrid Recurrent Neural Network-Generative Adversarial Network (RNN-GAN). Utilising the sequential data seen in images of lung tumours, the RNN-GAN architecture is used. In processing the sequential input, the RNN component looks for temporal relationships and patterns. The GAN component improves the training of the RNN for accurate classification by creating synthetic tumour specimens that resemble actual tumour images. In addition, the proposed approach pre-process lung tumour images using a Gaussian filter to improve their quality. The Gaussian filter improves feature extraction and the visibility of tumour borders by reducing noise and smoothing the pictures. The proposed experimental findings on a dataset of lung tumours shows that the suggested strategy successful. In comparison to conventional techniques, the hybrid RNN-GAN delivers higher accuracy in lung tumour identification due to the incorporation of the Gaussian filter. While the GAN component creates realistic tumour samples for improved training, the RNN component efficiently captures the sequential patterns of tumour images. The lung tumour images are pre-processed using a Gaussian filter, which greatly enhances image quality and facilitates precise feature extraction. The proposed hybrid RNN-GAN with the Gaussian filter shows promising potential for accurate and early detection of lung tumours. The integration of deep learning techniques with image pre-processing methods can contribute to the advancement of lung cancer diagnosis and treatment, ultimately improving patient outcomes and survival rates. Further research and validation are necessary to explore the full potential of this approach and its applicability in clinical settings.

Author 1: Atul Tiwari

Author 2: Shaikh Abdul Hannan

Author 3: Rajasekhar Pinnamaneni

Author 4: Abdul Rahman Mohammed Al-Ansari

Author 5: Yousef A.Baker El-Ebiary

Author 6: S. Prema

Author 7: R. Manikandan

Keywords: Lung tumour; recurrent neural network; generative adversarial network; CT images; hybrid

Paper 70: U-Net-based Pancreas Tumor Segmentation from Abdominal CT Images

Abstract: There is no doubt that pancreatic cancer is one of the most deadly types of cancer. In order to diagnose and stage pancreatic tumors, computed tomography (CT) is widely used. However, manual segmentation of volumetric CT scans is a time-consuming and subjective process. It has been shown that the U-Net model is highly effective for semantic segmentation, although several deep learning models have been proposed. In this study, we propose a U-Net-based method for pancreatic tumor segmentation from abdominal CT images and demonstrate its simplicity and effectiveness. Using the U-Net architecture, the pancreas is segmented from CT slices in the first stage, while tumors are segmented from masked CT images in the second stage. For validation set of NIH dataset and according to the proposed method's dice scores, the pancreas segmentation and tumor segmentation performance was outstanding, demonstrating its potential to identify pancreatic cancer efficiently and accurately.

Author 1: H S Saraswathi

Author 2: Mohamed Rafi

Keywords: U-net; deep learning; segmentation; computed tomography images; hyper parameters; PDAC

Paper 71: Personating GA Neural Fuzzy Hybrid System for Computing HD Probability

Abstract: The cardiovascular disease (CD) is a widespread, dangerous sickness involving an excessive rate of demise that necessitates quick piousness for care and cure. There are numerous diagnostic methods, such as angiography, available to diagnose heart disease (HD). ML is an extremely leading option for scientists for discovering prediction-based explanations for heart disease, and several machine learning algorithms are discovered to find the leading key results in community assistance. Researchers are presented with numerous conventional approaches, and various supportive algorithmic sequences formulated through the artificial neural network (NN) family, such as adaptive, convolutional, and de-convolutional NN, and various extended versions of hybrid combinations, originate with suitable outcomes. This research integrated the design and computational analysis of a unified model through a genetic algorithm-based Neural Fuzzy Hybrid System, which is formulated for CD prediction. It included a dual hybrid model to forecast CD and measure the degree of a healthy heart, as well as more precise heart attack complications. Stage 1 of the study's implications integrates the two stages and plans HD prediction using patient data. The input was processed in stages. First, the data was delivered in pre-processing mode. Next, the mRMR algorithm was used to select features. Finally, the model was trained using a variety of ML algorithms, including SVM, KNN, NB, DT, RF, LR, and NN. The results were compared, and based on those findings, the model was tuned to produce the best results. In stage 2, HA possibilities and occurrences are determined by FuzIS intelligence using data from the first stage, which includes more than 13000 pre-generated rules of fuzzy implications. These rules cover both normal-level and dangerous-level cases, and the medical parameters are integrated and tuned to produce membership functions that are then sent to the model. It is composed with the comparison of a unified system, which consists of Genetic algorithms, Neural networks, and Fuzzy Inference systems. In the experiment, gaussian MF sketched the continuous series of data, enabling the inference system to generate a good accuracy of 94% in calculating the problem probability.

Author 1: Rahul Kumar Jha

Author 2: Santosh Kumar Henge

Author 3: Sanjeev Kumar Mandal

Author 4: C Menaka

Author 5: Deepak Mehta

Author 6: Aditya Upadhyay

Author 7: Ashok Kumar Saini

Author 8: Neha Mishra

Keywords: Dickey-Fuller test case (DF-TC); HA prediction (HAP); heart rate variability (HRV); artificial based neural network (AbNN); Fuzzy Inference System (FuzIS); genetic-based algorithm (GbA); multi-objective evolutionary Fuzzy classifier (MOEFC); heart attack (HA); fuzzification-mode (FuzM); de-fuzzification-mode (De-FuzM)

Paper 72: DeepCyberDetect: Hybrid AI for Counterfeit Currency Detection with GAN-CNN-RNN using African Buffalo Optimization

Abstract: Modern technology has made a big contribution to the distribution of counterfeit money and the valuation of it. This paper recommends a deep learning-based methodology for currency recognition in order to extract attributes and identify money values; machine learning's binary classification task of fake currency detection. One can train a model that can distinguish between real and fake banknotes if one has enough information about actual and fake notes. The vast majority of older systems relied on hardware and techniques for image processing. Using such strategies renders identifying fake currency more challenging and inefficient. The proposed system has suggested deploying a deep convolution neural network to figure out fake currency in order to solve the aforementioned issue. By analyzing the images of the currency, our technique finds counterfeit notes. The transfer-learned convolutional neural network is trained using data sets that represent 2000 different currency notes in order to learn the unique characteristics map of the currencies. After becoming familiar with the feature map, the network is capable of real-time phoney cash detection. It is surprising how well deep learning models perform in photo classification tasks. The Deep CNN model that has been created in the proposed approach helps in the detection of the fake note without really manually extracting the properties of photographs. The model trains from the data set produced during training, letting us to identify fake currency. In multiple instances, techniques for deep learning have been shown to be more effective. Thus, deep learning is used to boost currency recognition accuracy. Among the techniques used are the African Buffalo Optimization Approach (ABO), recurrent neural networks (RNN), convolutional neural networks, generative adversarial networks (GAN) for identifying bogus notes, and classical neural networks.

Author 1: Franciskus Antonius

Author 2: Jarubula Ramu

Author 3: P. Sasikala

Author 4: J. C. Sekhar

Author 5: S. Suma Christal Mary

Keywords: Fake currency; convolutional neural network; generative adversarial networks; recurrent neural network; African Buffalo Optimization

Paper 73: Prediction of Cardiac Arrest by the Hybrid Approach of Soft Computing and Machine Learning

Abstract: Cardiac-related diseases are the major reason for the increased mortality rate. The early predictions of cardiac diseases like ventricular fibrillation (VF) are always challenging for doctors and data analysts. Early prediction of these diseases can save million lives. If the symptoms of these diseases are predicted early, the chance of survival increases significantly. For the prediction of Ventricular fibrillation (VF), several researchers have used Heart Rate Variability Analysis (HRV); various alternatives by combining the features taken from several areas to explore the prediction outcome. Several techniques like spectral Analysis, Rough Set Theory (RST), Support Vector Machine (SVM), and Adaboost techniques have not required any pre-processing. In this work, randomly medical-related data sets are taken from various parts of Odisha, applying regression and Rough Set techniques, reducing the dimension of the data set. Application of Rough Set Theory (RST) on the data set is not only useful in dimension reduction but also gives a set of various alternatives. This work's last section uses a comparative analysis between AdaBoost combined with RST and Empirical mode decomposition (EMD).

Author 1: Subrata Kumar Nayak

Author 2: Sateesh Kumar Pradhan

Author 3: Sujogya Mishra

Author 4: Sipali Pradhan

Author 5: P. K. Pattnaik

Keywords: Ventricular fibrillation (VF); heart rate variability (HRV); Rough Set Theory (RST); support vector machine (SVM); regression analysis; Adaboost method

Paper 74: An Improved Artificial Bee Colony Optimization Algorithm for Test Suite Minimization

Abstract: Software testing is essential process for maintaining the quality of software. Due to changes in customer demands or industry, software needs to be updated regularly. Therefore software becomes more complex and test suite size also increases exponentially. As a result, testing incurs a large overhead in terms of time, resources, and costs associated with testing. Additionally, handling and operating huge test suites can be cumbersome and inefficient, often resulting in duplication of effort and redundant test coverage. Test suite minimization strategy can help in resolving this issue. Test suite reduction is an efficient method for increasing the overall efficacy of a test suite and removing obsolete test cases. The paper demonstrates an improved artificial bee colony optimization algorithm for test suite minimization. The exploitation behavior of algorithm is improved by amalgamating the teaching learning based optimization technique. Second, the learner performance factor is used to explore the more solutions. The aim of the algorithm is to remove the redundant test cases, while still ensuring effectiveness of fault detection capability. The algorithm compared against three established methods (GA, ABC, and TLBO) using a benchmark dataset. The experiment results show that proposed algorithm reduction rate more than 50% with negligible loss in fault detection capability. The results obtained through empirical analysis show that the suggested algorithm has surpassed the other algorithms in performance.

Author 1: Neeru Ahuja

Author 2: Pradeep Kumar Bhatia

Keywords: Test suite; test suite minimization; TLBO; ABC; nature inspired algorithm

Paper 75: Automated Characterization of Autism Spectrum Disorder using Combined Functional and Structural MRI Analysis

Abstract: Autism Spectrum Disorders (ASD) are among the most critical health concerns of our time. These disorders typically present challenges in social interaction, communication, and exhibit repetitive behaviors. To diagnose and customize medical treatments for ASD effectively, the development of robust neuroimaging biomarkers is indispensable. Although extensive studies have recently delved into this area, only a handful have explored the differences between ASD and NC. This study aspires to shed light on this relationship by analyzing both structural and functional brain data associated with ASD. We aim to provide an extensive characterization of ASD by combining techniques of structural and functional analysis. The framework we propose is based on analyzing the differences in structural and functional aspects between ASD and development control (DC) subjects. The study leverages a substantial dataset of 1114 T1-weighted structural and functional Magnetic Resonance Imaging comprising 521 individuals with ASD and 593 controls, ranging in age from 5 to 64 years. These subjects are divided into three broad age categories. Utilizing automated labeling, we compute the features from subcortical and cortical regions. Statistical analyses help identify disparities between ASD and DC subjects. Principal Component Analysis (PCA) is employed to select the most discriminative features, which are subsequently used for classifying the two groups via an Artificial Neural Network (ANN) analysis. Our preliminary findings reveal a significant difference in the distribution of all tested features and subcortical regions between ASD subjects and DC subjects. Through our work, we contribute towards an enhanced understanding of ASD, potentially paving the way for future research and therapeutic interventions.

Author 1: Nour El Houda Mezrioui

Author 2: Kamel Aloui

Author 3: Amine Nait-Ali

Author 4: Mohamed Saber Naceur

Keywords: Autism spectrum disorder (ASD); Magnetic Resonance Imaging (MRI); functional Magnetic Resonance Imaging (fMRI); Artificial Neural Network (ANN)

Paper 76: Blockchain Architecture Based on Decentralised PoW Algorithm

Abstract: Blockchain has gained increasing popularity across various industries due to its decentralized, stable, and secure nature. Consensus algorithms play a crucial role in maintaining the security and efficiency of Blockchain systems and selecting the right algorithm can lead to significant performance improvements. This article aims to provide a comparative review of the most used Blockchain consensus algorithms, highlighting their strengths and weaknesses. Additionally, we propose a dissociated architecture for an efficient Blockchain system that doesn't compromise on security. A comparison is made between this architecture and the reviewed algorithms, considering aspects such as algorithm performance, energy consumption, mining, decentralization level, and vulnerability to security threats. The research findings demonstrate that the proposed architecture can support complex algorithms with high security while addressing issues related to efficiency, processing performance, and energy consumption.

Author 1: Cinthia P. Pascual Caceres

Author 2: Jose Vicente Berna Martinez

Author 3: Francisco Maciá Pérez

Author 4: Iren Lorenzo Fonseca

Author 5: Maria E. Almaral Martinez

Keywords: Blockchain technology; proof of work; consensus algorithm; proof of stake; Dissociated-PoW; security; performance

Paper 77: Sentiment Analysis of Code-mixed Social Media Data on Philippine UAQTE using Fine-tuned mBERT Model

Abstract: The Universal Access to Quality Tertiary Education (UAQTE) marks a significant policy change in the Philippines. While the program’s objective is to offer free higher education and tertiary education subsidies to eligible Filipino students, its viability and effectiveness have been subject to scrutiny and continuous evaluation. This study explores the sentiments of Filipinos towards UAQTE. Leveraging a fine-tuned multilingual Bidirectional Encoder Representations from Transformers (mBERT) model, we conducted sentiment analysis on code-mixed data. With minimal preprocessing, our model achieved an accuracy of 80.21% and an F1 score of 81.14%, surpassing previous related studies and confirming its effectiveness in handling code-mixed data. The results reveal that the majority of social media users view UAQTE positively or beneficially. However, negative sentiments highlight concerns related to subsidy delays, alleged fund misuse, and application challenges. Additionally, neutral sentiments center around subsidy-related announcements. These findings provide valuable insights for its key stakeholders involved in the implementation, enhancement, and evaluation of UAQTE.

Author 1: Lany L. Maceda

Author 2: Arlene A. Satuito

Author 3: Mideth B. Abisado

Keywords: Sentiment analysis; UAQTE; code-mixing; policy-making; multilingual BERT

Paper 78: Criminal Law Risk Management and Prediction Method based on Echo State Network

Abstract: Criminal law plays an important role in maintaining social security and achieving effective social control. However, criminal law has hidden risks that cannot be ignored at the legislative, judicial and theoretical levels. This paper starts from all aspects of criminal law, analyzes criminal law risk and its management measures, and predicts and analyzes criminal law risk through echo state network model. The prediction results of the echo state network model fit well with the actual situation, and its verification can provide reference for the study of criminal law risk prediction and management systems. Legislative risk and theoretical risk belong to social factors and are also fundamental risks of criminal law. Judicial risk is mainly manifested in the level of judicial power. Criminal law is closely related to the political environment, social system, economic system, etc. In criminal law legislation, we should pay attention to the balance between criminal law rules and realistic social functions, and properly control social risks, so as to avoid the criminal law risks brought by the establishment of risky criminal law, and provide the necessary guarantee for the national security system.

Author 1: Zhe Li

Keywords: Echo state network; model; criminal law; risk prediction; risk prediction

Paper 79: A Novel 2D Deep Convolutional Neural Network for Multimodal Document Categorization

Abstract: Digitized documents are increasingly becoming prevalent in various industries. The ability to accurately classify these documents is critical for efficient and effective management. However, digitized documents often come in different formats, making document classification a challenging task. In this paper, we propose a multimodal deep learning approach for digitized document classification. The proposed approach combines both text and image modalities to improve classification accuracy. The model architecture consists of a convolutional neural network (CNN) for image processing and a recurrent neural network (RNN) for text processing. The output features from the two modalities are then merged using a fusion layer to generate the final classification result. The proposed approach is evaluated on a dataset of digitized documents from various industries, including finance, healthcare, and legal fields. The experimental results demonstrate that the multimodal approach outperforms single-modality approaches, achieving high accuracy for document classification. The proposed model has significant potential for applications in various industries that rely heavily on document management systems. For example, in the finance industry, the proposed model can be used to classify loan applications or financial statements. In the healthcare industry, the model can classify patient records, medical images, and other medical documents. In the legal industry, the model can classify legal documents, contracts, and court filings. Overall, the proposed multimodal deep learning approach can significantly improve document classification accuracy, thus enhancing the efficiency and effectiveness of document management systems.

Author 1: Rustam Abkrakhmanov

Author 2: Aruzhan Elubaeva

Author 3: Tursinbay Turymbetov

Author 4: Venera Nakhipova

Author 5: Shynar Turmaganbetova

Author 6: Zhanseri Ikram

Keywords: Scanned documents; classification; document categorization; artificial intelligence; machine learning; deep learning

Paper 80: Artificial Neural Network for Binary and Multiclassification of Network Attacks

Abstract: Diving into the complex realm of network security, the research paper investigates the potential of leveraging artificial neural networks (ANNs) to identify and classify network intrusions. Balancing two distinct paradigms – binary and multiclassification – the study breaks fresh ground in this intricate field. Binary classification takes the stage initially, offering a bifurcated outlook: network traffic is either under attack, or it's not. This lays the foundation for an intuitive understanding of the network landscape. Then, the spotlight shifts to the finer-grained multiclassification, navigating through a realm that holds five unique classes: Normal traffic, DoS (Denial of Service), Probe, Privilege, and Access attacks. Each class serves a specific function, ranging from harmless communication (Normal) to various degrees and kinds of malicious intrusion. By integrating these two approaches, the research illuminates a path towards a more comprehensive understanding of network attack scenarios. It highlights the role of ANNs in enhancing the precision of network intrusion detection systems, contributing to the broader field of cybersecurity. The findings underline the potency of ANNs, offering fresh insights into their application and raising questions that promise to push the frontiers of cybersecurity research even further.

Author 1: Bauyrzhan Omarov

Author 2: Alma Kostangeldinova

Author 3: Lyailya Tukenova

Author 4: Gulsara Mambetaliyeva

Author 5: Almira Madiyarova

Author 6: Beibut Amirgaliyev

Author 7: Bakhytzhan Kulambayev

Keywords: Neural networks; artificial intelligence; detection; classification; attacks; network security

Paper 81: Lung Nodule Segmentation and Classification using U-Net and Efficient-Net

Abstract: The ability to detect lung cancer has led to better health outcomes. Deep learning techniques are widely used in the medical field to detect lung tumors at an early stage. Deep learning models such as U-Net, Efficient-Net, Resnet, VGG-16, etc. have been incorporated in various studies to detect lung cancer accurately. To enhance the detection performance, this work proposes an algorithm that combines U-Net and Efficient-Net neural networks for lung nodule segmentation and classification. A feature-extraction-based semi-supervised method is used to take advantage of the huge amount of CT scan images with no pathological labels. Semi-supervised learning is achieved using a feature pyramid network (FPN) with ResNet-50 model for feature extraction and a neural network classifier for predicting unlabelled nodules. The main innovation of U-Net is the skip-connections, which give the decoder access to the features that the encoder learned at various scales and enable accurate localization of lung nodules. Efficient-Net uses depth, width, and resolution scaling, combined with a compound coefficient that uniformly scales all network dimensions, resulting in an efficient neural network for image classification. This work has been evaluated on the publicly available LIDC-IDRI dataset and outperforms most existing methods. The proposed algorithm aims to address issues such as a high false-positive rate, small nodules, and a wide range of non-uniform longitudinal data. Experiment results show this model has a higher accuracy of 91.67% when compared with previous works.

Author 1: Suriyavarman S

Author 2: Arockia Xavier Annie R

Keywords: Cancer; CT; U-Net; efficient-net; feature; accuracy

Paper 82: Providing an Improved Resource Management Approach for Healthcare Big Data Processing in Cloud Computing Environment

Abstract: Due to the gathering of big data and the advancement of machine learning, the healthcare industry has recently experienced fast change. Acceleration of operations related to the analysis and retrieval of healthcare data is essential to facilitate surveillance. However, providing healthcare to the community is a complex task that is highly dependent on data processing. Also, processing health metadata can be very expensive for organizations. To meet the strict service quality requirements of the healthcare industry, large-scale healthcare data processing in the cloud confederation has emerged as a viable option. However, there are many challenges, including optimal resource management for metadata processing. Based on this, in the present study, a fuzzy solution for determining the optimal cloud using the resource forecasting technique is presented for health big data processing. During job processing, a fuzzy selection-based VM migration technique was used to move a virtual machine (VM) from a high-load server to a low-load server. The proposed architecture is divided into regional and global levels. After evaluating the local component, requests are sent to the global component. If the local component cannot meet the requirements, the request is sent to the global component. The hierarchical structure of the proposed method requires the generation of delivered requests before estimating the available resources. The proposed solution is compared with PSO and ACO algorithms according to different criteria. The simulation results show the effectiveness and efficiency of the model compared to alternative methods.

Author 1: Fei Zhou

Author 2: Huaibao Ding

Author 3: Xiaomei Ding

Keywords: Healthcare; big data; cloud confederation; service quality; Cloud Resource Management (CRM)

Paper 83: Open Information Extraction Methodology for a New Curated Biomedical Literature Dataset

Abstract: The research articles contain a wealth of information about the interactions between biomedical entities. However, manual relation extraction processing from the literature by domain experts takes a lot of time and money. In addition, it is often prohibitively expensive and labor-intensive, especially in biomedicine where domain knowledge is required. For this reason, computer strategies that can use unlabeled data to lessen the load of manual annotation are of great relevance in biomedical relation extraction. The present study solves relation extraction tasks in a completely unsupervised scenario. This article presents an unsupervised model for relation extraction between medical entities from PubMed abstracts, after filtration and preprocessing the abstracts. The verbs and relationship types are embedded in a vector space, and each verb is mapped to the relation type with the highest similarity score. The model achieves competitive performance compared to supervised systems on the evaluation using ChemProt and DDI datasets, with an F1-score of 85.8 and 88.5 respectively. These improved results demonstrate the effectiveness of extracting relations without the need for manual annotation or human intervention.

Author 1: Nesma Abdel Aziz Hassan

Author 2: Rania Ahmed Abdel Azeem Abul Seoud

Author 3: Dina Ahmed Salem

Keywords: Relation extraction; BERT; open information extraction; biomedical literature; ChemProt; DDI

Paper 84: Detection and Investigation Model for the Hard Disk Drive Attacks using FTK Imager

Abstract: A computer hard disk drive (HDD) is a device that stores, organizes, and manages computer data. In general, it is used for system storage, in which the computer maintains its operating system and other programs. A hard disk drive can, however, be physically damaged as well as affected by software errors, data corruption, and viruses that are used by attackers to cause damage. This study aims to develop a detection and investigation model (DIM) for HDD to detect and investigate HDD attacks using the FTK Imager forensic tool. The design science method is adapted to develop and evaluate the DIM. The developed DIM consists of three main phases: detection, gathering, and analysis. In order to evaluate the capabilities of the developed DIM for HDD, a real scenario was used. According to the results, the DIM can detect and investigate the HDD easily using FTK Imager. Thus, organizations can use the developed DIM to detect, investigate, mitigate, or avoid HDD threats.

Author 1: Ahmad Alshammari

Keywords: HDD; cybercrimes; design science method; digital forensic tools; FTK imager

Paper 85: Application of VR Technology Based on Gesture Recognition in Animation-form Capture

Abstract: To accurately capture the posture of animation characters in virtual vision and optimize the user's experience when wearing virtual vision equipment, the hybrid Gaussian model has gained wide attention. However, various types of animation show an exponential growth trend, and the hybrid Gaussian model is prone to low-dimensional explosion when processing these single frames. Based on the mixed Gaussian model, this study conducts animation character gesture recognition experiments on the Disert data set to solve these problems. Meanwhile, it is improved by frame rate reduction method to generate fusion algorithm. In this paper, the video is first grayened and filtered, and the model feature points of the image are marked. Then the weight learning rate is introduced and added to the set of pixels, and then the peak signal-to-noise ratio of Wronsky function is adjusted by changing the parameters. Then similar image sets are extracted and the structure elements are opened and closed. Finally, the proposed algorithm is applied to Disert data set. Meanwhile, the prediction accuracy of PSO is tested and compared with fusion algorithm. A total of 400 experiments were conducted, and the prediction accuracy of the fusion algorithm reached 392 times, with an accuracy of 98.0%. The accuracy of PSO is close to that of fusion algorithm (88.2%). It is verified that the suggested model can identify the four common gestures of cartoon characters well, and users will get a good viewing experience.

Author 1: Jing Yang

Author 2: Hao Zhang

Keywords: Frame rate reduction method; model feature points; Wronsky function; mixed Gaussian model; weight learning rate

Paper 86: GDM-PREP: A Rule-Based Technique to Enhance Early Detection of Gestational Diabetes Mellitus

Abstract: Gestational diabetes mellitus (GDM), a condition occurring solely during pregnancy, poses risks to both expectant mothers and their infants, particularly among individuals with pre-existing risk factors. However, early diagnosis and effective management of GDM can help mitigate potential complications. As part of the Ministry of Health's efforts to enhance screening and management strategies for GDM in Malaysia, this study aims utilizing a rule-based technique, acting as an Expert System for Initial Screening of Gestational Diabetes Mellitus Detection. This application will facilitate early diagnosis by assessing risk factors and symptoms to calculate the probability of GDM occurrence and classify it as low, medium, or high. Functionality and usability tests are conducted to ensure error-free performance and gather user feedback. The study's findings indicate that the self-check GDM system effectively utilizes the algorithm, while the mobile application showcases good usability, achieving an above-average System Usability Scale (SUS) score.

Author 1: Ayunnie Azmi

Author 2: Nurulhuda Zainuddin

Author 3: Azmi Aminordin

Author 4: Masurah Mohamad

Keywords: Gestational diabetes mellitus (GDM); rule based; expert systems; risk factor

Paper 87: Innovating Art with Augmented Reality: A New Dimension in Body Painting

Abstract: This study investigates the fusion of augmented reality (AR) and body painting as a novel concept for artistic expression. By combining the immersive capabilities of AR with the creative potential of body painting, this research explores individuals' perceptions and attitudes towards this innovative artistic approach from an HCI perspective. Drawing upon the Technology Acceptance Model (TAM) and the Diffusion of Innovation Theory (DIT), the study examines the factors influencing individuals' acceptance and intention to engage in AR-integrated body painting. Additionally, the research explores the mediating role of artistic expression in understanding the impact of these factors on the actual outcomes of this merged concept. A sample of 212 respondents participated in an online survey to accomplish the research objectives. The survey comprehensively measured participants' perceptions of innovativeness, social system support, perceived usefulness, perceived ease of use, artistic expression, and behavioral intention towards AR-integrated body painting. Rigorous data analysis was conducted using Partial Least Squares Structural Equation Modeling (PLS-SEM) to examine the intricate relationships between the variables. The findings underscore the significant impact of factors such as Innovativeness, social system support, perceived usefulness, and perceived ease of use on individuals' acceptance and intention to engage in AR-integrated body painting from an HCI perspective. Moreover, the study reveals the mediating role of artistic expression in connecting these influential factors with the actual outcomes of this merged concept. These empirical insights substantially contribute to our understanding of the fundamental mechanisms driving the adoption and utilization of AR in artistic practices, particularly within the domain of body painting, from both an artistic and HCI standpoint.

Author 1: Dou Lei

Author 2: Wan Samiati Andriana W. Mohamad Daud

Keywords: Augmented reality; body paintings; artistic expression; technology acceptance

Paper 88: The Essence of Software Engineering Framework-based Model for an Agile Software Development Method

Abstract: Agile development's rapid growth is due to its ability to address complex problems and facilitate a smooth transition from traditional methods. However, no single Agile method can fit every organization, which leads to a lack of adoption guidelines. It triggers this investigation by proposing an Agile development method model based on the Essence of software engineering framework and incorporating the common ground of popular methods such as Scrum, Kanban, Extreme programming, SAFe, Less, Nexus, Spotify Agile, Scrum of Scrums, and Disciplined Agile. The Essence of software engineering framework provides an approach for organizations to develop software development methods based on common ground or shared understanding among methods. We enhance this approach for Agile methods, resulting in a model to support organizations in developing their Agile methods and practices. Moreover, Design Science Research (DSR) was employed as a methodology to construct the artifact, demonstration, and evaluation. We demonstrated the model in an Agile product development at a national-wide bank in Indonesia. This investigation enhances Agile methods in SWEBOK's Software Engineering Models and Methods knowledge area, benefiting academics and practitioners. Practitioners can use the model as a reference to implement their Agile projects.

Author 1: Teguh Raharjo

Author 2: Betty Purwandari

Author 3: Eko K. Budiardjo

Author 4: Rina Yuniarti

Keywords: Agile; common ground; the essence of software engineering framework; Design Science Research (DSR)

Paper 89: Effective Face Recognition using Adaptive Multi-scale Transformer-based Resnet with Optimal Pattern Extraction

Abstract: Human face is the major characteristic for identifying a person and it helps to differentiate each person. Face recognition methods are mainly useful for determining a person’s identity with the help of biometric techniques. Face recognition methods are used in many practical applications like criminal identification, the phone unlocks systems and home security systems. It does not need any key and card, and it only requires facial images to provide high security over several applications. The interdependencies of the encryption methods are highly reduced in the deep learning-enabled face recognition models. Conventional methods did not satisfy the present demand due to poor recognition accuracy. Therefore, an advanced deep learning-based face recognition framework is implemented to authenticate the identity of individuals with high accuracy by using facial images. The required facial images are taken from the standard databases. The collected images are preprocessed using median filtering. The preprocessed facial images are subjected to spatial feature extraction, where the Local Binary patterns (LBP) and Local Vector Patterns (LVP) are utilized to extract the relevant optimal patterns from the facial images. Here, optimal pattern extraction is done with the Improved Rat Swarm Optimization Algorithm (IRSO). Then, the facial recognition is done over the extracted optimal features with the usage of the implemented Adaptive Multi-scale transformer-based Resnet (AMT-ResNet), where the parameters in the recognition network are optimized by using the IRSO. The efficiency of the developed deep learning adopted face recognition model is validated through different heuristics algorithms, and baseline face recognition approaches.

Author 1: Santhosh Shivaprakash

Author 2: Sannangi Viswaradhya Rajashekararadhya

Keywords: Face recognition; facial images; optimal pattern extraction rate; local binary patterns; local vector patterns; improved rat swarm optimization algorithm; adaptive multi-scale transformer-based Resnet

Paper 90: An Intelligent Malware Classification Model Based on Image Transformation

Abstract: Due to financial incentives, the number of malware infections is steadily rising. Accuracy and effectiveness are essential because malware detection systems serve as the first line of defense against harmful attacks. A zero-day vulnerability is a hole in the target operating system, device driver, application, or other tools employing a computer environment that was previously unknown to anybody other than the hacker. Traditional malware detection systems usually use conventional machine learning algorithms, which call for time-consuming and error-prone feature gathering and extraction. Convolutional neural networks (CNNs) have been demonstrated to outperform conventional learning techniques in a number of applications, including the classification of images. This success prompts us to suggest a CNN-based malware categorization architecture. We evaluated our methodology using a bigger dataset made up of 25 families within a corpus of 9342 malware. Last but not least, comparisons are made between the model's measurement and performance with other cutting-edge deep learning techniques. The overall testing accuracy of 98.31% in the provided results attested to the excellent accuracy and robustness of the suggested procedure at a lower computational cost.

Author 1: Mohamed Abo Rizka

Author 2: Mohamed Hamed

Author 3: Hatem A. Khater

Keywords: Malware Classification; zero-day; Convolutional Neural Networks (CNN); grayscale image transformation; Bytehist

Paper 91: Unsupervised Document Binarization of Engineering Drawings via Multi Noise CycleGAN

Abstract: The task of document binarization of degraded complex documents is tremendously challenging due to the various forms of noise often present in these documents. While the current state-of-the-art deep learning approaches are capable for the removal of various noise types in documents with high accuracy, they employ a supervised learning scheme which requires matching clean and noisy document image pairs which are difficult and costly to obtain for complex documents such as engineering drawings. In this paper, we propose our method for document binarization of engineering drawings using ‘Multi Noise CycleGAN’. The method utilizing unsupervised learning using adversarial and cycle-consistency loss is trained on unpaired noisy document images of various noise and image conditions. Experimental results for the removal of various noise types demonstrated that the method is able to reliably produce a clean image for any given noisy image and in certain noisy images achieve significant improvements over existing methods.

Author 1: Luqman Hakim Rosli

Author 2: Yew Kwang Hooi

Author 3: Ong Kai Bin

Keywords: Image processing and computer vision; generative adversarial networks; document binarization; deep learning

Paper 92: Light Weight Circular Error Learning Algorithm (CELA) for Secure Data Communication Protocol in IoT-Cloud Systems

Abstract: The data driven smart applications, utilize the IoT, Cloud Computing, AI and other digital technologies to create, curate and operate on large amounts of data to provide intelligent solutions for day-to-day problems. Security of Data in the IoT-Cloud systems has become very crucial as there are several attacks such as ransomware, data thieving, and data corruption, causing huge loss to the application users. The basic impediment in providing strong security solutions for the IoT systems, is due the resource limitations of IoT devices. Recently, there is an additional threat of quantum computing being able to break the traditional cryptographic techniques. The objective of this research is to address the bifold challenge and design a light weight quantum secure communication protocol for the IoT Cloud ecosystem. The Ring Learning With Errors (RLWE) lattice based cryptography has emerged as the most popular in the NIST PQC Standardization Program. A light weight Circular Learning Error Algorithm (CELA) has been proposed by optimizing RLWE to make it suitable for IoT-Cloud environment. The CELA inherits the advantages of quantum security and homomorphic encryption from RLWE. It is observed that CELA is light weight in terms of execution time and a slightly bigger cipher text size provides higher security as compared to RLWE. The paper also offers plausible solutions for future quantum secure cryptographic protocols.

Author 1: Mangala N

Author 2: Eswara Reddy B

Author 3: Venugopal K R

Keywords: Quantum secure cryptography; homomorphic encryption; lattice-based cryptography; Learning With Errors (LWE); Ring Learning With Errors (RLWE); Circular Error Learning Algorithm (CELA)

Paper 93: Intelligent Anomaly Detection Method of Gateway Electrical Energy Metering Devices using Deep Learning

Abstract: Accurate anomaly detection of gateway electrical energy metering device is important for maintenance and operations in the power systems. Traditionally, anomaly detection was typically performed manually through the analysis of the collected energy information. However, the manual process is time-consuming and labor-intensive. In this condition, this paper proposes a hybrid deep-learning model, which integrates Stacked Autoencoder (SAE) and Long Short-Term Memory (LSTM), for intelligently detecting the abnormal events of gateway electrical energy metering device. The proposed model named SAE-LSTM model, first uses SAE to extract deep latent features of three-phase voltage data collected from the gateway electrical energy metering device, and then adopts LSTM for separating the abnormal events based on the extracted deep latent features. The SAE-LSTM model, can effectively highlight the temporal information of the electrical data, thereby enhancing the accuracy of anomaly detection. The simulation experiments verify the advantages of the SAE-LSTM model in anomaly detection under different signal-to-noise ratios. The experimental results of real datasets demonstrate that it is suitable for anomaly detection of gateway electrical energy metering devices in practical scenarios.

Author 1: Lihua Zhang

Author 2: Xu Chen

Author 3: Chao Zhang

Author 4: Lingxuan Zhang

Author 5: Binghang Zou

Keywords: Anomaly detection; gateway electric energy metering device; stacked autoencoder; long short-term memory

Paper 94: Semantic Privacy Inference Preservation Algorithm for Indoor Trajectory

Abstract: Indoor location services have become an increasingly important part of our everyday lives in recent years. Despite the numerous benefits these services offer, serious concerns have arisen about the privacy of users’ locations. Adversaries can monitor user-requested locations in order to obtain sensitive information such as shopping patterns. Many users of indoor spaces want their movements and locations to be kept private so as not to reveal their visit to a particular zone inside buildings. Research on semantic indoor trajectory-based human movement data has primarily focused on finding routes without taking into account the protection of privacy. Hence, the servers on which trajectory data is stored are not completely secure. In this paper, we propose a semantic privacy inference preservation algorithm for an indoor trajectory that can issue path finding and navigation instructions while achieving good privacy protection of moving entities by generating ambiguous trajectory. The simulation of the proposed semantic indoor privacy algorithm was implemented in MATLAB.

Author 1: Abdullah Alamri

Keywords: Privacy; semantic ontology; indoor space; routing algorithm; spatial databases

Paper 95: Knee Cartilage Segmentation using Improved U-Net

Abstract: Patello-femoral joint stability is a complex problem and requires detailed anatomic parametric study for knowing the associated breakdowns of knee cartilage. Osteoarthritis is one of the main disorders, which disrupt the normal bio-mechanics and stability of the patello-femoral joint and for diagnosing osteoarthritis radiologists needs a lot of time to diagnose it. An improved network called PSU-Net is proposed for the automatic segmentation of femoral, tibia, and patella cartilage in knee MR images. The model utilizes a Squeeze and Excitation block with residual connection for effective feature learning that helps in learning imbalance anatomical structure between background, bone areas and cartilage. The severity of knee cartilage is measured through the Kellgren and Lawrence (KL) grading system by radiologists. Also, updated weighted loss function is used during training to optimize the model and improve cartilage segmentation. Results demonstrate that PSU-Net can accurately and quickly identify cartilages compared to the traditional procedures, aiding in the treatment planning in a very short amount of time. Future work will involve the use of augmentation methods and also use this architecture as a generator model for generative adversarial network to improve performance further. The utility of this work will help in analyzing the anatomy of the human knee by the radiologists in short amount of time that may prove helpful to standardize and automate patello-femoral measurements in diverse patient populations.

Author 1: Nawaf Waqas

Author 2: Sairul Izwan Safie

Author 3: Kushsairy Abdul Kadir

Author 4: Sheroz Khan

Keywords: Knee image segmentation; U-Net; loss function; squeeze and excitation

Paper 96: An Integrated Framework for Relevance Classification of Trending Topics in Arabic Tweets

Abstract: Social media platforms such as Twitter are a valuable source of information about current events and trends. Trending topics aim to promote public events such as political events, market changes, and other types of breaking news. However, with so much data being generated, it would be difficult to identify relevant tweets that are related to a particular trending topic. Therefore, in this paper, an integrated framework is proposed for the detection of the degree of relevance between Arabic tweets and trending topics. This framework integrates natural language processing, data augmentation, and machine learning techniques to identify text that is likely to be relevant to a given trending topic. The proposed framework was evaluated using a real-life dataset of Arabic tweets that was collected and labeled. The results of the evaluation showed that the proposed framework achieved the highest macro F1 score of 82% in binary classification (relevant/irrelevant) and 77% in categorical classification (degree of relevance), which outperforms the current state of the art.

Author 1: Abdullah M. Alkadri

Author 2: Abeer ElKorany

Author 3: Cherry A. Ezzat

Keywords: Trending topics; social media platforms; machine learning; Arabic relevance classification; data augmentation

Paper 97: Generative Adversarial Network-based Approach for Automated Generation of Adversarial Attacks Against a Deep-Learning based XSS Attack Detection Model

Abstract: Cross Site Scripting attack (XSS) is one of the most famous and dangerous web attacks. In XSS attacks, illegitimate technical methods are used by attackers to disclose sensitive data from web site users, which result in an important finance and reputation loss to the web site’s owner. There exist numerous XSS attack countermeasures. Deep Learning has been shown to be effective when used to detect XSS attacks in HTTP web requests. Yet, Deep Learning models are inherently vulnerable to adversarial attacks, which aim to deceive the detection model into mis-classifying malicious HTTP web requests. Thus, it is important to evaluate the robustness of the detection model against adversarial attacks before its deployment to production in real web applications. In this work, we developed a Generative Adversarial Network (GAN) model for automated generation of adversarial XSS attacks against an LSTM-based XSS attack detection model. We showed that the detection model performance drops drastically when evaluated on the XSS instances, originally used in the model development, but modified by the GAN model. We also provided some guidelines to the development of detection models that can defend against adversarial attacks in the particular context of web attacks detection.

Author 1: Rokia Lamrani Alaoui

Author 2: El Habib Nfaoui

Keywords: Deep learning; generative adversarial network; LSTM; web attacks; adversarial attacks; Cross Site Scripting attack

Paper 98: A Review on Machine-Learning and Nature-Inspired Algorithms for Genome Assembly

Abstract: Genome assembly plays a crucial role in the field of bioinformatics, as current sequencing technologies are unable to sequence an entire genome at once where the need for fragmenting into short sequences and reassembling them. The genomes often contain repetitive sequences and duplicated regions, which can lead to ambiguities during assembly. Thus, the process of reconstructing a complete genome from a set of reads necessitates the use of efficient assembly programs. Over time, as genome sequencing technology has advanced, the methods for genome assembly have also evolved, resulting in the utilization of various genome assemblers. Many artificial intelligence techniques such as machine learning and nature-inspired algorithms have been applied in genome assembly in recent years. These technologies have the potential to significantly enhance the accuracy of genome assembly, leading to functionally correct genome reconstructions. This review paper aims to provide an overview of the genome assembly, highlighting the significance of different methods used in machine learning techniques and nature-inspiring algorithms in achieving accurate and efficient genome assembly. By examining the advancements and possibilities brought about by different machine learning and metaheuristics approaches, this review paper offers insights into the future directions of genome assembly.

Author 1: Asmae Yassine

Author 2: Mohammed Essaid Riffi

Keywords: Artificial intelligence; genome assembly; machine learning; bioinformatics; bio-inspired algorithms

Paper 99: Predicting Customer Segment Changes to Enhance Customer Retention: A Case Study for Online Retail using Machine Learning

Abstract: In today’s highly competitive marketplace, advertisers strive to tailor their messages to specific individuals or groups, often overlooking their most significant clients. The Pareto principle, asserting that 80% of sales come from 20% of customers, offers valuable insights, imagine if companies could accurately forecast this vital 20% and recognize its historical significance. Predicting customer lifetime value (CLV) at this juncture becomes crucial in aiding firms to effectively prioritize their efforts. To achieve this, organizations can leverage predictive models and analytical tools to target specific customers with tailored campaigns, enabling well-informed decisions about advertising investments. By being aware of these segment transitions, advertisers can efficiently deploy resources and increase their return on investment. By implementing the strategies outlined in this study, businesses can gain a competitive edge by identifying and retaining their most valuable clients. The potential for growth and client retention is immense when anticipating changes in customer segments and adjusting advertising strategies accordingly. This paper provides a comprehensive methodology, tools, and insights to assist marketers in optimizing their advertising campaigns by anticipating customer lifetime value and actively predicting changes in client segmentation.

Author 1: Lahcen ABIDAR

Author 2: Dounia ZAIDOUNI

Author 3: Ikram EL ASRI

Author 4: Abdeslam ENNOUAARY

Keywords: Customer segment changes; customer retention; marketing actions; informed decisions; advertising strategies

Paper 100: Review of Existing Datasets Used for Software Effort Estimation

Abstract: The Software Effort Estimation (SEE) tool calculates an estimate of the amount of work that will be necessary to effectively finish the project. Managers usually want to know how hard a new project will be ahead of time so they can divide their limited resources in a fair way. In fact, it is common to use effort datasets to train a prediction model that can predict how much work a project will take. To train a good estimator, you need enough data, but most data owners don’t want to share their closed source project effort data because they are worried about privacy. This means that we can only get a small amount of effort data. The purpose of this research was to evaluate the quality of 15 datasets that have been widely utilized in studies of software project estimation. The analysis shows that most of the chosen studies use artificial neural networks (ANN) as ML models, NASA as datasets, and the mean magnitude of relative error (MMRE) as a measure of accuracy. In more cases, ANN and support vector machine (SVM) have done better than other ML techniques.

Author 1: Mizanur Rahman

Author 2: Teresa Goncalves

Author 3: Hasan Sarwar

Keywords: Software effort estimation; software effort prediction; software effort estimation datasets

Paper 101: Employee Attrition Prediction using Nested Ensemble Learning Techniques

Abstract: In many industries, including the IT industry, rising employee attrition is a major concern. Hiring a candidate for an unsuitable job because of issues with the employment process can lead to employee attrition. Thus, enhancing the employment process would reduce the attrition rate. This paper aims to investigate the effect of ensemble learning techniques on enhancing the employment process by predicting employee attrition. This paper applied a two-layer nested ensemble model to the IBM HR Analytics Employee Attrition & Performance dataset. The performance of this model was compared to that of the random forest (RF) algorithm as a baseline for comparison. The results showed that the proposed model outperformed the baseline algorithm. The RF model achieved an accuracy of 94.2417%, an F1-score of 94.2%, and an AUC of 98.4%. However, the proposed model had the highest performance. It outperformed with an accuracy of 94.5255%, an F1-score of 94.5%, and an AUC of 98.5%. The performance of the proposed model was compared with that of the baseline comparison algorithm by using a paired t-test. According to the paired t-test, the performance of the proposed model was statistically better than that of the baseline comparison algorithm at the significance level of 0.05. Thus, the two-layer nested ensemble model improved the employee attrition prediction.

Author 1: Muneera Saad Alshiddy

Author 2: Bader Nasser Aljaber

Keywords: Nested ensemble learning; employee attrition; machine learning; employment process

Paper 102: Bird Detection and Species Classification: Using YOLOv5 and Deep Transfer Learning Models

Abstract: Bird detection and species classification are important tasks in ecological research and bird conservation efforts. The study aims to address the challenges of accurately identifying bird species in images, which plays a crucial role in various fields such as environmental monitoring, and wildlife conservation. This article presents a comprehensive study on bird detection and species classification using the YOLOv5 object detection algorithm and deep transfer learning models. The objective is to develop an efficient and accurate system for identifying bird species in images. The YOLOv5 model is utilized for robust bird detection, enabling the localization of birds within images. Deep transfer learning (TL) models, including VGG19, Inception V3, and EfficientNetB3, are employed for species classification, leveraging their pre-trained weights and learned features. The experimental findings show that the proposed approach is effective, with excellent accuracy in both bird detection and tasks for species classification. The study showcases the potential of combining YOLOv5 with deep transfer learning models for comprehensive bird analysis, opening avenues for automated bird monitoring, ecological research, and conservation efforts. Furthermore, the study investigated the effects of optimization algorithms, including SGD, Adam, and Adamax, on the performance of the models. The findings contribute to the advancement of bird recognition systems and provide insights into the performance and suitability of various deep transfer learning architectures for avian image analysis.

Author 1: Hoang-Tu Vo

Author 2: Nhon Nguyen Thien

Author 3: Kheo Chau Mui

Keywords: Bird detection; species classification; YOLOv5; deep transfer learning models; automated bird monitoring

Paper 103: FCCC: Forest Cover Change Calculator User Interface for Identifying Fire Incidents in Forest Region using Satellite Data

Abstract: For the ecosystem to maintain a balance between the social and environmental spheres, forests play a crucial role. The greatest threat to forests for this significance, is fires and natural disasters caused by several factors. It is crucial to assess the genesis and behavioral characteristics of fires in forest areas. The discovery of the forest fire areas and the intensity of the fire affected are greatly facilitated by the satellite image obtained by different sensors and data sets. We are suggesting a novel approach to compute changes using spectral indices, using landsat-9 and sentinel-2 satellite datasets for measuring the change in forest areas affected by fire accidents over Kochi areas on March 2023. Kochi is a city in Kerala, South India, and is located at 9° 50’ 20.7348” N and 77° 10’ 13.8828” E. coordinates. Computation is performed by calculating forest area before the fire incident (pre-fire) and after the fire incident (post-fire) and total loss is calculated by the difference between pre-fire and post-fire incident. The proposed work uses Sentinel-2 and Landsat-9 satellite images to recover burn scars using several vegetation indicators. We have identified the fire locations using the object-based classification approach. For verification of results computed by vegetation indices, we have also performed land use land cover classification and calculated the changes in forest areas. Accuracy is computed by the confusion matrix with an accuracy of 89.45% and the kappa coefficient with an accuracy of 87.68%. In particular, there was a strong correlation between forest loss and the burned area in the subtropical evergreen broadleaf forest zone (6.9%) and the deciduous coniferous forest zone (18.9%of the lands). These findings serve as a foundation for future forecasts of fire-induced forest loss in regions with similar climatic and environmental conditions.

Author 1: Anubhava Srivastava

Author 2: Sandhya Umrao

Author 3: Susham Biswas

Author 4: Rakesh Dubey

Author 5: Md. Iltaf Zafar

Keywords: GEE; remote sensing; classification; landsat; sentinel; forest fire

Paper 104: Enhancing Computer-assisted Bone Fractures Diagnosis in Musculoskeletal Radiographs Based on Generative Adversarial Networks

Abstract: Computer-Assisted Bone Fractures Diagnosis in musculoskeletal radiographs plays a crucial role in aiding medical professionals in accurate and timely fracture detection. In this work, we explore a Generative Adversarial Network based approach for this task, which is a powerful deep learning model capable of generating realistic images and detecting anomalies. Our proposed approach leverages the potential of GANs to generate synthetic radiographs with fractures and identify anomalous patterns, thereby enhancing fracture diagnosis. Through extensive experimentation and evaluation on musculoskeletal radiograph datasets (MURA), we demonstrate the effectiveness of GAN-based models in improving fracture detection performance by adopting several evaluation metrics notably accuracy, precision, F1-score and detection speed. These findings highlight the potential of integrating GANs into computer-assisted diagnosis, contributing to the advancement of fracture diagnosis methodologies in orthopedics. It is important to note that GANs operate by training a generator network to produce synthetic images and a discriminator network to distinguish between real and generated images. This adversarial process fosters the generation of realistic radiographs with fractures, enabling accurate and automated detection. Our findings contribute to the advancement of fracture diagnosis methodologies and pave the way for more efficient and precise diagnostic tools in the field of orthopedics.

Author 1: Nabila Ounasser

Author 2: Maryem Rhanoui

Author 3: Mounia Mikram

Author 4: Bouchra El Asri

Keywords: Deep learning; generative adversarial network; diagnosis; orthopedics; fracture detection; x-ray image

Paper 105: Performance Evaluation of Face Mask Detection for Real-Time Implementation on an Rpi

Abstract: Mask-wearing remains to be one of the primary protective measures against COVID-19. To address the difficulty of manual compliance monitoring, face mask detection models considerate of both frontal and angled faces were developed. This study aimed to test the performance of the said models in classifying multi-face images and upon running on a Raspberry Pi device. The accuracies and inference speeds were measured and compared when inferencing images with one, two, and three faces and on the desktop and the Raspberry Pi. With an increasing number of faces in an image, the models’ accuracies were observed to decline, while their speeds were not significantly affected. Moreover, the YOLOv5 Small model was regarded to be potentially the best model for use on lower resource platforms, as it experienced a 3.33% increase in accuracy and recorded the least inference time of two seconds per image among the models.

Author 1: Ivan George L. Tarun

Author 2: Vidal Wyatt M. Lopez

Author 3: Pamela Anne C. Serrano

Author 4: Patricia Angela R. Abu

Author 5: Rosula S.J. Reyes

Author 6: Ma. Regina Justina E. Estuar

Keywords: Face mask detection; multi-face detection; Raspberry Pi; embedded platform

Paper 106: A Hybrid TF-IDF and RNN Model for Multi-label Classification of the Deep and Dark Web

Abstract: The classification of content on the deep and dark web has been a topic of interest for researchers. Researchers focus on adopting more efficient and effective classification methods as the data available on deep and dark web platforms continues to grow. Multi-label classification is the approach for simultaneously categorizing content into multiple classes. To address this, a hybrid approach combining Term Frequency-Inverse Document Frequency (TF-IDF) and Recurrent Neural Network (RNN) has been proposed. The approach involves preprocessing a dataset of Hypertext Markup Language (HTML) documents, selecting specific HTML tags to generate embeddings using TF-IDF, and using an RNN model for multi-label classification. The proposed model was evaluated against commonly used methods (Binary Relevance, Classifier Chains, and Label Powerset) using precision, recall, and F1-score as evaluation metrics, demonstrating promising results in accurately classifying data from the deep and dark web. This contribution represents a noteworthy advancement for researchers and analysts working in this field.

Author 1: Ashwini Dalvi

Author 2: Soham Bhoir

Author 3: Nishavak Naik

Author 4: Atharva Kitkaru

Author 5: Irfan Siddavatam

Author 6: Sunil Bhirud

Keywords: Deep web; dark web; multi-label classification; TF-IDF; FastText; RNN

Paper 107: Smart-Agri: A Smart Agricultural Management with IoT-ML-Blockchain Integrated Framework

Abstract: This paper presents intuitive directions for field research by introducing a ground-breaking IoT-ML-driven intelligent farm management platform. This study’s main goal is to address agricultural difficulties by providing a thorough, integrated solution. This work makes a variety of important contributions. By utilizing cutting-edge technology like IoT and Machine Learning (ML), it first improves conventional farm management procedures. Farmers now have the capacity to remotely monitor and regulate irrigation management thanks to sensor-based real-time data. Second, based on data gathered from agricultural fields, our machine learning model offers improved water control management and fertilizer use recommendations, maximizing production while minimizing resource usage. The suggested solution also uses blockchain technology to create a safe, decentralized network that guarantees data integrity and defends against threats. We also introduce energy harvesting technology to address the issue of continuous energy supply for IoT devices, which lessens the load on farmers by removing the requirement for additional batteries. We achieved 89.5% accuracy in our proposed machine learning model. The suggested model would provide a variety of services to farmers, including pesticide recommendations and water motor control via mobile applications and a cloud database.

Author 1: Md. Mamun Hossain

Author 2: Md. Ashiqur Rahman

Author 3: Sudipto Chaki

Author 4: Humayra Ahmed

Author 5: Ahsanul Haque

Author 6: Iffat Tamanna

Author 7: Sweety Lima

Author 8: Most. Jannatul Ferdous

Author 9: Md. Saifur Rahman

Keywords: Smart agriculture; machine learning; internet of things; energy harvesting; blockchain technology

Paper 108: New Approach based on Association Rules for Building and Optimizing OLAP Cubes on Graphs

Abstract: The expansion of data has prompted the creation of various NoSQL (Not only SQL) databases, including graph-oriented databases, which provide an understandable abstraction for modeling complex domains and managing highly connected data. However, to add graph data to existing decision support systems, new data warehouse systems that consider the special characteristics of graphs need to be developed. This work proposes a novel method for creating a data warehouse under a graph database and demonstrates how OLAP (Online Analytical Processing) structures created for reporting can be handled by graph databases. Additionally, the paper suggests using aggregation algorithms based association rules techniques to improve the efficiency of reporting and data analysis within a graph-based data warehouse. Finally, we provide a Cypher language implementation of the suggested approach to evaluate and validate our approach.

Author 1: Redouane LABZIOUI

Author 2: Khadija LETRACHE

Author 3: Mohammed RAMDANI

Keywords: NoSQL; graph-oriented databases; data warehouse; OLAP; aggregation algorithms; association rules; cypher language

Paper 109: Cloud Task Scheduling using Particle Swarm Optimization and Capuchin Search Algorithms

Abstract: Cloud providers offer heterogeneous virtual machines for the execution of a variety of tasks requested by users. These virtual machines are managed by the cloud provider, eliminating the need for users to set up and maintain their hardware. This makes accessing the computing resources necessary to run applications and services more accessible and cost-effective. The task scheduling problem can be expressed as a discrete optimization issue known as NP-hard. To address this problem, we propose a hybrid meta-heuristic algorithm using the Capuchin Search Algorithm (CapSA) and the Particle Swarm Optimization (PSO) algorithm. PSO excels in global exploration, while CapSA is adept at fine-tuning solutions through local search. We aim to achieve better convergence and solution quality by integrating both algorithms. Our proposed method's performance is thoroughly evaluated through extensive experimentation, comparing it to standalone PSO and CapSA approaches. The findings reveal that our hybrid algorithm outperforms the individual techniques in terms of both total execution time and total execution cost metrics. The novelty of our work lies in the synergistic integration of PSO and CapSA, addressing the limitations of traditional optimization methods for cloud task scheduling. The proposed hybrid approach opens up intriguing directions for future research in dynamic task scheduling, multi-objective optimization, adaptive algorithms, integration with emerging technologies, and real-world deployment scenarios.

Author 1: Gang WANG

Author 2: Jiayin FENG

Author 3: Dongyan JIA

Author 4: Jinling SONG

Author 5: Guolin LI

Keywords: Cloud computing; virtualization; task scheduling; optimization; resource utilization; capuchin search algorithm; particle swarm optimization

Paper 110: A Modified Hybrid Algorithm Approach for Solving Harmonic Problems in Power Systems

Abstract: A fundamental problem with electrical systems' power quality is electrical harmonics. In order to limit harmonics and their effects on power systems, filters used in electric power systems must be designed with the consideration of power harmonics. The study's suggested approach differs from other hybrid strategies that have been previously published, and the processes that are expected mostly center on cutting down on computing complexity and time. The voltage and current waveforms of distribution networks have started to be significantly distorted over the past 20 years due to the increased use of power electronic equipment and non-linear loads. This paper provides a new hybrid approach for harmonic estimation. Harmonic estimation of these deformed waveforms is a nonlinear problem because sinusoidal waveforms contain nonlinear distortions. As a result, the corresponding combined technique splits the problem of harmonic estimation into two independent problems due to the slow convergence of nonlinear problems in the estimate of harmonic components. The algorithm used in this study first estimates amplitude and frequency using the fuzzy logic control (FLC) approach, a non-linear estimator. The objective function is then to minimize the error value for both the original signal and the estimated signal using the genetic algorithm, a non-linear estimator. The experiments show that the proposed method for determining harmonic estimation time is 36% better than comparable methods. As a result, the suggested technique offers a number of benefits, including very quick computation times, more precise evaluation of amplitude and phase values for all conditions, and complexity in the outcomes.

Author 1: Ning WANG

Author 2: Qiuju DENG

Keywords: Harmonic estimation; fuzzy PD controlled; harmonic components

Paper 111: Inverted Ant Colony Optimization Algorithm for Data Replication in Cloud Computing

Abstract: Data replication is crucial in enhancing data availability and reducing access latency in cloud computing. This paper presents a dynamic duplicate management method for cloud storage systems based on the Inverted Ant Colony Optimization (IACO) algorithm and a fuzzy logic system. The proposed approach optimizes data replication decisions focusing on energy consumption, response time, and cost. Extensive simulations demonstrate that the IACO-based method outperforms existing techniques, achieving a remarkable 25% reduction in energy consumption, a significant 15% improvement in response time, and a substantial 20% cost reduction. By addressing the research gap concerning integrating IACO and fuzzy logic for data replication, our work contributes to advancing cloud computing solutions for large datasets. The proposed method offers a viable and efficient approach to improve resource utilization and system performance, benefiting various scientific fields.

Author 1: Min YANG

Keywords: Cloud computing; data replication; cloud data centers; reliability; energy efficiency; inverted ant colony optimization algorithm

Paper 112: Improved Cat Swarm Optimization Algorithm for Load Balancing in the Cloud Computing Environment

Abstract: Recently, cloud computing has gained recognition as a powerful tool for providing clients with flexible platforms, software services, and cost-effective infrastructures. Cloud computing is a form of distributed computing that allows users to store and process data in a virtual environment instead of a physical server. This is beneficial because it allows businesses to quickly scale up or down their computing capacity, reducing the need to invest in expensive hardware. As cloud tasks continue to grow exponentially and the usage of cloud services increases, scheduling these tasks across diverse virtual machines poses a challenging NP-hard optimization problem with substantial requirements, including optimal resource utilization levels, a short execution time, and a reasonable implementation cost. The issue has consequently been addressed using a variety of meta-heuristic algorithms. In this paper, we propose a new load-balancing approach using the Cat Swarm Optimization (CSO) algorithm in order to distribute the load among the various servers within a data center. Statistical analyses indicate that our algorithm is superior to previous research with regard to energy consumption, makespan, and time required up to 30%, 35%, and 40%, respectively.

Author 1: Wang Dou

Keywords: Cloud computing; resource utilization; load balancing; optimization

Paper 113: Automatic Fraud Detection in e-Commerce Transactions using Deep Reinforcement Learning and Artificial Neural Networks

Abstract: Fraud is a serious issue that has plagued e-commerce for many years, and despite significant efforts to combat it, current fraud detection strategies only catch a small portion of fraudulent transactions. This results in substantial financial losses, with billions of dollars being lost each year. Given the expected surge in the volume of online transactions in the upcoming years, there is a critical need for improved fraud detection strategies. To tackle this problem, the article proposes a deep reinforcement learning approach for the automatic detection of fraudulent e-commerce transactions. The architecture's policy is built on the implementation of artificial neural networks (ANNs). The classification problem is viewed as a step-by-step decision-making procedure. The implementation of the model involves the use of the artificial bee colony (ABC) algorithm to acquire initial weight values. After that, in each step, the agent obtains a sample and performs a classification, with the environment providing a reward for each classification action. To encourage the model to concentrate on detecting fraudulent transactions precisely, the reward for identifying the minority class is higher than that for the majority class. With the aid of a supportive learning setting and a specific reward system, the agent ultimately determines the best approach to achieve its objectives. The performance of the suggested model is assessed utilizing a publicly available dataset contributed by the Machine Learning group at the Université Libre de Bruxelles. The experimental outcomes, determined using recognized evaluation measures, indicate that the model has attained a high level of accuracy. As a result, the suggested model is considered appropriate for identifying deceitful transactions in e-commerce.

Author 1: Yuanyuan Tang

Keywords: Fraud detection; reinforcement learning; artificial neural network; artificial bee colony; imbalanced classification

Paper 114: Optimal Scheduling using Advanced Cat Swarm Optimization Algorithm to Improve Performance in Fog Computing

Abstract: Fog computing can be considered a decentralized computing approach that essentially extends the capabilities offered by cloud computing to the periphery of the network. In addition, due to its proximity to the user, fog computing proves to be highly efficient in minimizing the volume of data that needs to be transmitted, reducing overall network traffic, and shortening the distance that data must travel. But this technology, like other new technologies, has challenges, and scheduling and optimal allocation of resources is one of the most important of these challenges. Accordingly, this research aims to propose an optimal solution for efficient scheduling within the fog computing environment through the application of the advanced cat swarm optimization algorithm. In this solution, the two main behaviors of cats are implemented in the form of seek and tracking states. Accordingly, processing nodes are periodically examined and categorized based on the number of available resources; servers with highly available resources are prioritized in the scheduling process for efficient scheduling. Subsequently, the congested servers, which may be experiencing various issues, are migrated to alternative servers with ample resources using the virtual machine live migration technique. Ultimately, the effectiveness of the proposed solution is assessed using the iFogSim simulator, demonstrating notable reductions in execution time and energy consumption. So, the proposed solution has led to a 20% reduction in execution time while also improving energy efficiency by more than 15% on average. This optimization represents a trade-off between improving performance and reducing resource consumption.

Author 1: Xiaoyan Huo

Author 2: Xuemei Wang

Keywords: Scheduling; fog computing; optimal balancing; cat swarm optimization algorithm

Paper 115: A Hybrid Cryptography Method using Extended Letters in Arabic and Persian Language

Abstract: Cryptography is widely used in information security systems. In encryption, the goal is to hide information in such a way that only the sender and receiver are aware of the existence of communication and information. Encryption takes place in various media, such as image, sound and text. Today, the rapid growth of network technologies and digital tools has made digital delivery fast and easy. However, the distribution of digital data in public networks such as the Internet has various challenges due to copyright infringement, forgery, coding and fraud. Therefore, methods of protecting digital data, especially sensitive data, are very necessary. Accordingly, in this article, a combined solution is used based on the technique of stretching the letters and making minor changes in the letters that have closed spaces so that the bits related to the hidden text can be inserted into a Persian or Arabic language. For this purpose, a new solution has been designed, in which the cover text is similar to the normal text, with the difference that, in addition to the extended letters that are longer due to the status of the secret message, it also has some prepositions, which have spaces. They are empty and closed. Of course, this difference in the closed space between the original letters and the changed letters will be very slight, and as a result, there will not be much difference between them that the normal user can feel the change. Finally, the proposed solution has been evaluated with the help of the MATLAB program, and according to the rate parameter of encryption capacity, the results show that the proposed method has an average encryption capacity of more than 50% compared to other common solutions.

Author 1: Ke Wang

Keywords: Cryptography; extended letters; Persian language; Arabic language

Paper 116: A Comprehensive Review of Fault-Tolerant Routing Mechanisms for the Internet of Things

Abstract: The Internet of Things (IoT) facilitates intelligent communication and real-time data collection through dynamic networks. The IoT technology is ideally suited to meet intelligent city requirements and enable remote access. Several cloud-based approaches have been proposed for constrained IoT systems, including scalable data storage and effective routing. In real-world scenarios, the effectiveness of many methods for wireless networks and communication links can be challenged due to their unpredictable characteristics. These challenges can result in path failures and increased resource utilization. To enhance the reliability and resilience of IoT networks in the face of failures, fault tolerance mechanisms are crucial. Network failures can occur for various reasons, including the breakdown of the wireless nodes' communication module, node failures caused by battery drain, and changes in the network topology. Addressing these issues is essential to ensure the continuous and reliable operation of IoT networks. Fault-tolerant routing plays a critical role in IoT-based networks, but no systematic and comprehensive research has been conducted in this area. Therefore, this paper aims to fill this gap by reviewing state-of-the-art mechanisms. An analysis of the practical techniques leads to recommendations for further research.

Author 1: Zhengxin Lan

Keywords: Internet of things; routing; data transmission; fault-tolerant; review

Paper 117: DeepShield: A Hybrid Deep Learning Approach for Effective Network Intrusion Detection

Abstract: In today's rapidly evolving digital landscape, ensuring the security of networks and systems has become more crucial than ever before. The ever-present threat of hackers and intruders attempting to disrupt networks and compromise online services highlights the pressing need for robust security measures. With the continuous advancement of security systems, new dangers arise, but so do innovative solutions. One such solution is the implementation of Network Intrusion Detection Systems (NIDSs), which play a pivotal role in identifying potential threats to computer systems by categorizing network traffic. However, the effectiveness of an intrusion detection system lies in its ability to prepare network data and identify critical attributes necessary for constructing robust classifiers. In light of this, this paper proposes, DeepShield, a cutting-edge NIDS that harnesses the power of deep learning and leverages a hybrid feature selection approach for optimal performance. DeepShield consists of three essential steps: hybrid feature selection, rule assessment, and detection. By combining the strengths of machine learning and deep learning technologies, a new solution is developed that excels in detecting network intrusions. The process begins by capturing packets from the network, which are then carefully preprocessed to reduce their size while retaining essential information. These refined data packets are then fed into a deep learning algorithm, which employs machine learning characteristics to learn and test potential intrusion patterns. Simulation results demonstrate the superiority of DeepShield over previous approaches. NIDS achieves an exceptional level of accuracy in detecting malicious attacks, as evidenced by its outstanding performance on the widely recognized CSE-CIC-DS2018 dataset.

Author 1: Hongjie Lin

Keywords: Network intrusion detection system; IDS; cyber security; machine learning; deep learning

Paper 118: Deep-Learning-based Analysis of the Patterns Associated with the Changes in the Grit Scores and Understanding Levels of Students

Abstract: The purpose of this study is to classify the pattern of the understanding level changes for university students during class term, and analyze the relation between them and the changes in the Grid score before and after the class term. Dynamic time warping was applied for classification of the understanding level, and the decision tree was applied to analyze the relation between the changes in the understanding level and that in the Grid score. As a result, it was shown that a large variety of the patterns of changes in the understanding level, and the relations between the understanding level and Grid score cover a wide variety, too. It is necessary to take these results for conducting effective lectures.

Author 1: Ayako OHSHIRO

Keywords: Time series; dynamic time warping; decision tree; Grit