		FeedBack \| Contact Us \| Links \| Site Map

Welcome to Journal of Theoretical and Applied Information Technology

Submit Paper / Call for Papers

Journal receives papers in continuous flow and we will consider articles from a wide range of Information Technology disciplines encompassing the most basic research to the most innovative technologies. Please submit your papers electronically to our submission system at http://jatit.org/submit_paper.php in an MSWord, Pdf or compatible format so that they may be evaluated for publication in the upcoming issue. This journal uses a blinded review process; please remember to include all your personal identifiable information in the manuscript before submitting it for review, we will edit the necessary information at our side. Submissions to JATIT should be full research / review papers (properly indicated below main title).

Journal of Theoretical and Applied Information Technology
May 2024 | Vol. 102 No.9

Title:	ENERGY EFFICIENT ROUTING IN QUANTUM FLYING AD HOC NETWORK (Q-FANET) USING MAMDANI FUZZY INFERENCE ENHANCED DIJKSTRAS ALGORITHM (MFI-EDA)
Author:	S. P. GEETHA, N. MOHANA SORUBHA SUNDARI, J. RAMKUMAR, R. KARTHIKEYAN
Abstract:	Quantum Flying Ad Hoc Networks (Q-FANETs) present a unique paradigm for communication, leveraging quantum principles to enable secure and efficient data transmission. However, routing in Q-FANETs poses significant challenges due to dynamic topology changes and limited communication resources. This paper proposes a novel routing approach utilizing the Mamdani Fuzzy Inference Enhanced Dijkstras Algorithm (MFI-EDA) tailored for Q-FANET environments. The working mechanism of MFI-EDA involves integrating fuzzy logic with Dijkstras algorithm to intelligently adapt routing decisions based on environmental conditions, such as node mobility and energy levels, and network dynamics, such as link quality and traffic congestion. This hybrid approach enhances traditional routing algorithms by incorporating fuzzy logic to provide robustness and adaptability in Q-FANETs. The essential contribution lies in the seamless integration of fuzzy inference, which enables MFI-EDA to dynamically adjust routing paths based on real-time environmental feedback, resulting in improved energy efficiency and reliability. The performance of MFI-EDA in Q-FANET scenarios is evaluated through extensive simulation experiments, demonstrating its effectiveness in achieving energy-efficient and reliable routing. Results indicate that MFI-EDA outperforms traditional routing approaches, offering promising prospects for efficient communication in quantum-enabled ad hoc networks.
Keywords:	Quantum Network, Q-FANETs, Mamdani Fuzzy Inference, Dijkstras Algorithm, Routing
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	NEW METHOD FOR FINDING THE WEIGHT DISTRIBUTION AND SPECTRUM BY TESTING THE OPTIMAL SOLUTION OF THE POINTS IN PG(3,2)
Author:	HAMID MOHAMMED KHALAF , ASMAA SALAH ALDDIN SULAIMAN
Abstract:	In this paper, the basic solution was found by using Vogel's approximation method for the projective space PG(3,2) when mi=5,7,15 by converting the points and lines of PG(3,2) into a transportation problem with five rows and seven columns. The problem was balanced and treating Degeneracy. Then we tested the optimality of the basic solution using the modified distribution method. And then we construct a new linear codes of dimension p=4 and smallest length mi for which a [mi, p, d]-codes exists, and by a geometric method we found the weight distribution and the spectrum of a linear code [mi, p, d] over a field F2.
Keywords:	Vogel's Approximation, Projective Space, Linear Codes, Weight Distribution, Spectrum.
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	EXPLORING ARO-LOGISENT: DELVING INTO DATA COLLECTION FOR ADVANCED SENTIMENT ANALYSIS WITH AMAMI RABBIT OPTIMIZATION-BASED LOGISTIC REGRESSION
Author:	D. J. ANITHA MERLIN , D. VIMAL KUMAR
Abstract:	Sentiment analysis, a crucial component of natural language processing, aims to discern the underlying sentiment conveyed in textual data. This research explores the fusion of Amami Rabbit Optimization (ARO) with Logistic Regression (LR) to enhance sentiment analysis performance. ARO, inspired by the foraging behavior of Amami rabbits, offers a novel metaheuristic approach for optimizing model parameters, while LR provides a robust framework for sentiment classification. The proposed integration leverages the strengths of both methodologies to overcome challenges inherent in sentiment analysis, including feature selection, model training, and accuracy optimization. This study investigates the effectiveness of the ARO-LR hybrid approach through empirical experiments conducted on diverse datasets sourced from social media platforms and product reviews. Evaluation metrics such as precision, recall, F1-score, and accuracy are employed to assess the performance of the integrated model. Results indicate significant improvements in sentiment classification accuracy and robustness compared to traditional LR models. The findings highlight the efficacy of integrating metaheuristic optimization techniques with conventional machine learning algorithms for advancing sentiment analysis capabilities in real-world applications.
Keywords:	Sentiment, Reviews, Online Shopping, Classification, Amazon, Neural Network, Rabbit Optimization, Logistic Regression
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	SKIN LESION CLASSIFICATION FOR MELANOMA USING DEEP LEARNING
Author:	K SANDHYARANI KUNDRA , I V S VENUGOPAL , CH SITA KUMARI , ANUSHA GUTTI
Abstract:	Deep learning is a branch of machine learning and Artificial Intelligence that imitates the way people learn specific types of information. Image classification is an area in Computer Vision where a computer can analyze an image and identify or estimate the probability of the class or category the image falls under. Melanoma has become more common over the past 30 years, and early detection has a big effect on lowering death rates from this type of skin cancer. The existing system consists of several physical laboratory test reports that are analyzed by a doctor or a cancer expert to detect the presence of melanoma. This alone is insufficient in countries with huge population like India, where cancer hospitals and labs are few. The proposed system consists of an image classifier that takes a dermatoscopic image as input and predicts whether a person has melanoma or not. A reliable automated system that can tell whether melanoma is present in a dermatoscopic image of lesions is a very helpful tool for medical diagnosis. This system uses an ISIC dataset specially curated for melanoma analysis. Thus, the proposed system serves as an automated, expedient, and practical method for detecting melanoma from the image of a skin lesion.
Keywords:	Computer Vision, Deep Learning, Image Classification, Machine Learning, Transfer Learning
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	HYBRID FINGERPRINTING AND PEDESTRIAN DEAD RECKONING USING MACHINE LEARNING FOR INDOOR POSITIONING SYSTEM
Author:	ALFITO HARLIM, LIM , GEDE PUTRA KUSUMA
Abstract:	Indoor Positioning System is one of the hot research topics in the last years for it gives opportunity to be used in many business platform. BLE technology is used for indoor positioning system to reduce material and energy cost over time compared to other technologies which cost more. There has been a lot of study to increase positioning accuracy. One of the latest improvements is to use hybrid approach which combines the results of two methods. In this study, we propose a positioning algorithm for indoor positioning system using fingerprinting approach using two machine learning: Artificial Neural Network (ANN) and Support Vector Regression (SVR), and a hybrid of fingerprinting and Pedestrian Dead Reckoning (PDR) using ANN and SVR. Hybrid fingerprinting using Weighted K-Nearest Neighbor (W-KNN) and PDR is used as benchmark of this experiment. ANN and SVR are proposed as the machine learning used for both fingerprinting and hybrid method to combine with PDR. On benchmark, hybrid fingerprinting KNN using ANN achieve positioning root-mean-squared error 147.94 cm. The proposed hybrid fingerprinting ANN achieved 103.29 cm using ANN and 160.86 cm using SVR.
Keywords:	Indoor Positioning System, Fingerprinting Methods, Pedestrian Dead Reckoning, Hybrid Method, Machine Learning
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	IMPLEMENTATION OF ATTACK DETECTION AND MITIGATION FOR SECURING CLOUD BASED IOT NETWORK
Author:	ARCHANA D.WANKHADE , KISHOR P.WAGH
Abstract:	Now a days Internet of Things (IoT) has become fastest computing technology which makes human life easier and comfortable. IoT is important for smart system like homes, transportation, farming etc. Using this robust technology improves efficiency, mobility and cost reduction. But due to heterogeneous nature of this IoT network, it has many security issues. The devices in IoT network are generally attacked by intruders. Hence risk of security is high in IoT network than other computing paradigm. This is the reason why traditional security techniques are not useful in these IoT network. A holistic solution is required for fulfilling security requirement of IoT network. The existing security issues like authentication, access control, network security are not fulfilling challenges in large IoT system having number of smart devices. In this communication of Cloud based IoT network sensor data is transmitted from IoT network to cloud. This path is vulnerable. The objective of this work is to secure this path by proposed Attack Detection and Mitigation using ML approach for Cloud Internet of Things (IoT) network. For this proposed work, license software NetSim Standard v13.3 is used for identification of malicious node in Cloud based IoT network. The work was tested with ML approaches used for Attack Detection and Mitigation and efficiency is calculated to improve the performance. Performance of proposed model is improved using Decision Tree(Train Score 100%,Test Score 99.39% ) comparing with K-Nearest Neighbors (Train Score 98.52% and Test Score 98.24%) and Logistic Regression algorithms(Train Score 92.87% and Test Score 92.31%).
Keywords:	Attack Detection and Mitigation, IoT, KNN, Lightweight Cryptography, Machine Learning.
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	AN ENHANCED COMMUNITY DETECTION METHOD USING LABEL PROPAGATION ALGORITHM WITH ANT COLONY OPTIMIZATION TECHNIQUE
Author:	D. DHANALAKSHMI , DR. G. RAJENDRAN
Abstract:	The structure of a community is a crucial element in understanding complex networks. It provides valua-ble insights into both the arrangement and function of the network, aiding our comprehension of dynamic phenomena like epidemics and information propagation. While the Label Propagation Algorithm (LPA) is widely recognized for community detection due to its linear time complexity, it has a notable drawback. In comparison with other algorithms label propagation has advantages in its running time and amount of a priori information needed about the network structure. The biggest advantage of label propagation algorithm is that it owns excellent running time as well as simple algorithm process. LPA generates unstable community assignments, resulting in different combinations of communities with each execution on the same network. This unpredictability fosters instability and the emergence of large, less informative communities. To overcome these drawbacks, an Enhanced Community Detection approach, a combination of the Label Propagation Algorithm and Ant Colony Optimization (ECDLPA-ACO) tech-nique has been proposed in this paper. ECDLPA-ACO not only propagates labels but also optimizes modu-larity measures by clustering similar vertices based on local similarities within the network. Experimental results on established social network datasets showcase the superiority of ECDLPA-ACO over comparable community detection algorithms like Louvain Algorithm, Infomap Algorithm, and traditional Label Prop-agation Algorithm. ECDLPA-ACO outshines in scalability, average execution time, modularity, and com-putational efficiency.
Keywords:	Community Detection, Louvain Algorithm, Infomap Algorithm, Label Propagation Algorithm, Ant Colony Optimization technique
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	INTEGRATING AI FOR ENHANCED FAULT DETECTION IN INDUSTRIAL SYSTEMS: EVALUATING MACHINE AND DEEP LEARNING APPROACHES IN THE INDUSTRY 4.0 AND IIOT ERA
Author:	YOUSSEF ZERGUIT , YOUNES HAMMOUDI , MOSTAFA DERRHI
Abstract:	In the transformative landscape of Industry 4.0, the integration of Artificial Intelligence (AI) within the Industrial Internet of Things (IIoT) has emerged as a cornerstone for advancing operational efficiency and reliability. This paper explores the application of various AI methodologies, including both machine learning and deep learning approaches, to enhance fault detection in industrial systems, particularly focusing on three-phase electrical systems. Utilizing an integrated system architecture comprising Power Monitoring Units (PMUs) and advanced computational units, we implement and evaluate a suite of AI models such as Support Vector Machines (SVM), Decision Trees, K-Nearest Neighbors (KNN), Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU) networks. Our comprehensive analysis reveals the nuanced capabilities and performance metrics of these models in the context of real-time fault detection, thereby providing pivotal insights for deploying AI-driven diagnostics in industrial settings.
Keywords:	Three-Phase Systems, Industry 4.0, Industrial Internet Of Things (Iiot), Artificial Intelligence (Ai), Machine Learning, Deep Learning, Support Vector Machines (Svm), Decision Trees, K-Nearest Neighbors (Knn), Artificial Neural Networks (Ann), Convolutional Neural Networks (Cnn), Recurrent Neural Networks (Rnn), Long Short-Term Memory (Lstm), Gated Recurrent Units (Gru).
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	NOVEL ROUTING PROTOCOL TO OVERCOME PACKET DELAY IN MOBILE ADHOC NETWORK
Author:	DR.S.HEMALATHA, RABINARAYAN SETHI, M.RAJASEKARAN, RAVULA ARUN KUMAR, M. VIMALA5, VELMURUGAN V, SHAIK RAZIA, J.DEEPA
Abstract:	Packet delay in the wireless network leads the degrading the overall performance, many of the research work was done to overcome the packet delay in the nodes by inventing routing protocols with latest methods but the research remains the same stage. Existing routing strategy determines by using the dynamic manipulation of different parameters this is additional overhead to the routing protocol for determines the path. In this article concentrates on inventing a new routing protocol to overcome packet delay in Mobile Adhoc Network with a support of simple parameters called forwarding time of the each packet. To achieve this objective the Forward Time Based Routing Protocol (FTRP) introduces to monitoring the forwarded time of the every nodes present in the communication. The proposed forward time routing protocol was implemented with Network simulator and the simulation results are compared with existing methods of Proposed_TAODV, C-AODV, A-AODV and ML-AODV based routing protocol then the compared results are proved the packet delivery ratio is 90% to 94% and End to End Delay is 6.2% to 43.4 %. Simulation result in all the factors the proposed FTRP modes proved best result overall MANET the performance factors are excellent in 78%.
Keywords:	MANET, Attackers, Gray Hole Attackers, Forward Time Detection Technique, Forward Time
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	INTENT CLASSIFICATION FOR MALAYSIAN ACADEMIC WRITERS PROOFREADER CHATBOT USING MACHINE LEARNING
Author:	SITI NOOR BAINI MUSTAFA, LAILATUL QADRI BINTI ZAKARIA
Abstract:	The intent classification component is important in developing a chatbot as it helps the chatbot system to understand the meaning and purpose of conversation from the user. Earlier researchers have developed datasets for niche domains and analyzed various input representations and machine learning techniques for chatbot intent classification. However, there is no dataset and intent classification analysis for chatbot readily available in the niche of proofreading. Other than that, this study finds out the feature, input representation, and the best machine learning classifier that are suitable for intent classification analysis. This research is divided into seven main phases. The first phase is the feasibility study. The second phase is the dataset development. The third phase is the text preprocessing phase where input is cleaned and normalized. The fourth phase is the feature extraction phase whereby features are extracted using POS tagger, bag of words technique, and bigram words technique. The fifth phase is the input representation phase using Term Frequency – Inverse Document Frequency (TF-IDF) technique, Word2Vec embedding, or One-Hot encoding technique. Finally, the sixth phase is the intent classification phase using machine learning algorithms. The machine learning methods tested were Support Vector Classifier, Support Vector Machine, Stochastic Gradient Descent, and Naïve Bayes. The final phase is the testing phase. This study finds that the combination of noun and verb as features and using One-Hot encoding as input representation together with Support Vector Machine as the machine learning technique produces the best performing classifier for this research with 0.89 accuracy. This study hopes to pioneer the development of a proofreading chatbot that can help and take over a proofreaders task of answering the questions asked by Malaysian academic writers regarding the grammar corrections made.
Keywords:	Chatbot; Intent classification, Machine learning, Natural language processing; Proofreading corpus
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	UTILIZING GENETIC ALGORITHM INTEGRATED WITH INTELLIGENT OPERATORS AND SARSA FOR EXTRACTING HIGH UTILITY ITEMSETS
Author:	LOGESWARAN K, SAVITHA S, SURESH S3, ANANDAMURUGAN S
Abstract:	This research article presents a novel approach for mining High Utility Itemsets (HUIs) by integrating Genetic Algorithm (GA) with SARSA algorithm. It begins by providing a comprehensive overview of GA's fundamental principles and operational procedures, followed by an in-depth exploration of SARSA algorithm components, supported by diagrammatic representations. The core contribution of this study is the introduction of the Intelligent Genetic Algorithm with on-policy Reinforcement Learning (IGA_RLON) methodology, which is thoroughly elaborated upon. The effectiveness of IGA_RLON is meticulously evaluated in terms of execution time, convergence speed, and the percentage of successfully mined HUIs, through comparative analysis with established methods such as IGA_RLOFF, HUPEUMU-GRAM, and HUIM-BPSO. This article aims to advance the field of HUI mining by proposing a robust and efficient algorithmic framework.
Keywords:	Genetic Algorithm, SARSA Algorithm, Reinforcement Algorithm, High Utility Itemset Mining, Data Mining, Control Parameters
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	DEEP LEARNING FOR SENTIMENT BASED DYNAMIC STOCK SYMBOL ANALYSIS
Author:	PAUL JOSHI, SANTA MARIA, LAKSHMI K.S, DIVYA JAMES, LIYAN GRACE SHAJI
Abstract:	The financial markets are highly dynamic and constantly affected by a multitude of global events, ranging from political crises and global pandemics to trade competition, innovation, and scientific advancements. These disruptions reverberate through supply chains across the globe and impact the overall demand for goods and services. Research across various domains suggests that stock market behavior cannot be solely explained by predictable patterns or trends. Emotions play a significant role, influencing rational thinking and social behavior, thus making the stock market an embodiment of the collective social mood. In light of this perspective, analyzing public sentiment becomes crucial in predicting stock market movements. As the volume of textual data, such as news articles and tweets, continues to surge, they serve as valuable indicators of the prevailing public sentiment. Consequently, the proposed approach involves developing a sophisticated system capable of dynamically fetching recent tweets related to specific sectors and employing deep learning techniques, such as LSTM and GRU, for sentiment analysis. By leveraging these insights, investors can make informed decisions in the dynamic realm of financial markets.
Keywords:	Deep Learning, GRU, LSTM, Sentiment Analysis, Stock Market
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	RISK ASSESSMENT THREAT MODELLING USING AN INTEGRATED FRAMEWORK TO ENHANCE SECURITY
Author:	P. SUBHASH, MOHAMMED QAYYUM, K. MEHERNADH, K. JEEVAN SAHIT, C. LIKHITHA VARSHA, M. NEVAN HARDEEP
Abstract:	Today, in the digital world, the security of systems is crucial because, with continuous information exchange and huge quantities of data being processed, the protection of operation processes and data assets becomes paramount. Threat modeling is an important part of cybersecurity management methodology that looks for any possible threats or weaknesses that can break the system or endanger the environment. This paper is an attempt to analyze all the models of threat modeling by taking STRIDE, PASTA, DREAD, TREK, VAST and Attack Trees as references. An integrated model is suggested that combines the benefits of existing approaches, which includes the adoption of a comprehensive frame to deal with cyber threats. This methodology emphasizes iterative refinement and rigorous testing to ensure the effectiveness of threat mitigation strategies. By incorporating user-friendly web portals and the integration of new technologies, this framework enhances usability and addresses emerging threats. Overcoming the key issues in cybersecurity through integrated threat modeling is provided by combining approaches like STRIDE (security threat rating indicating, systemwide generic method), and PASTA (risk-focused, even in its extensive nature), DREAD, TREK, VAST and Attack Trees. This is due to the fact that the six methodologies and the ones that from Attack Trees offer a solution that is adaptable to various organizational contexts. An addition a unique triad is provided - asset-centric, attack-centric, and software-centric - to expand the protection coverage against the different kinds of threats and vulnerabilities. Proposed model Risk assessment threat modelling using an integrated framework to enhance security of action comprises the iterative revision and the exhaustive verification to guarantee the efficiency of prevention power. With the help of easy-to-use web portals and the integrated new technologies that make it all usable and design flexible, and timely as it faces new threats. Conducted on the basis of comparison; this analysis highlights the exclusive advantages of the design in the exact localization and mitigation of threats, achieving an accuracy of 94.2%. The integrated threat modeling framework that is presented in this essay represents a robust and dynamic approach toward cybersecurity, and it aims to improve the security and resilience of the system in the context of the continually changing threats in the cybersecurity world.
Keywords:	Threat Modelling, Integrated Framework, Iterative Refinement, Rigorous Testing, Web-portal
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	OPTIMIZING HEART DISEASE PREDICTION MODELS USING GENETIC ALGORITHMS: A METAHEURISTIC APPROACH
Author:	SUSHILA PALIWAL, SURAIYA PRAVEEN, M. AFSHAR ALAM, JAWED AHMED
Abstract:	Cardiovascular diseases, including heart disease, remain a significant global health concern. Early detection and accurate prediction of heart disease risk factors are crucial for effective prevention and intervention. This research paper uses a metaheuristic approach with genetic algorithms (GAs) to optimize hyperparameter settings to improve predictive accuracy of heart disease models. Using genetic algorithms as a metaheuristic approach, this research article aims to improve the accuracy and generalizability of heart disease prediction models. The primary objective is to converge towards optimal model parameters, ultimately maximizing accuracy or minimizing error. The study investigates the effectiveness of combining genetic algorithms with machine learning in improving heart disease prediction, with a focus on enhancing accuracy, generalizability, and scalability. Through comparison with traditional methods, the research assesses the superiority of the proposed approach and its potential contributions to heart disease diagnosis and treatment. Notably, the research stands out for its use of genetic algorithms in conjunction with cross-validation to optimize hyperparameters, identify optimal model parameters, and evaluate performance by minimizing errors. The application of the Support Vector Machine (SVM) classifier with optimized hyperparameters yielded a significant 97% improvement in accuracy with the heart disease dataset, surpassing results from previous studies. This research thus highlights the promise of genetic algorithms in enhancing heart disease prediction models and advancing healthcare analytics
Keywords:	Metaheuristic Approach, Genetic Algorithm, Hyperparameter Optimization, Support Vector Machine, Heart Disease Prediction
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	TRANSFORMING HEALTHCARE WITH FEDERATED LEARNING: SECURING FOG SYSTEMS FOR THE NEXT GENERATION
Author:	TAYSEER ALKHDOUR, AITIZAZ ALI, MOHAMMED ALMAIAH, TING TIN TIN, MOHMOOD A. AL-SHAREEDA, ROMMEL ALALI,THEYAZAN ALDAHYANI, ABDALWALI LUTFI
Abstract:	Recent advancements in fog computing, coupled with the Internet of Things (IoT) technology, encompass data analysis and artificial intelligence (AI) systems. Nonetheless, the inherent weakness of the current paradigm lies in its susceptibility to security risks and vulnerabilities. Security concerns and cyber-attacks remain significant challenges within fog computing environments. Collaborative attacks such as phishing assaults, along with replay attacks, exemplify common security threats. In this scenario, each layer the edge layer for sensing, the fog layer for processing, and the top layer encompassing storage and administration (cloud) - is vulnerable to attacks. The Internet of Things (IoT) in the fog (Fog-IoT) is widely acknowledged as the cornerstone of the contemporary world. Consequently, intelligent healthcare systems are increasingly prevalent. However, the rapid proliferation of IoT-based medical devices and technologies presents challenges in maintaining a comprehensive medical IoT system within budget constraints. While single Cloud Platforms (CP) would be immensely beneficial if standardized, achieving this through a decentralized fog computing system proves challenging. To address this, we propose a hybrid-deep learning protocol aimed at safeguarding electronic medical records from security breaches while simultaneously reducing latency. Additionally, we introduce scalable federated centered (FC) learning integrated with Blockchain-based data storage and retrieval. The proposed framework offers a secure, reliable, and low-latency approach to healthcare systems using a homomorphic distributed protocol.
Keywords:	EMR; IoMT; Cyber-risks; Sensors, Fog Computing; Cloud Computing; Security; Privacy.
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	THE EFFECT OF INFLUENCER REPUTATION ON DIGITAL ENGAGEMENT AND CONSUMER BRAND RELATIONSHIP
Author:	ZING-YUN ZENG, YISITIE XING, CHANG-HYUN JIN
Abstract:	The purpose of this study is to identify the sub-components of the value of an influencer to understand their relationship with digital integration. Influencer value components are divided into attractiveness, reliability, and expertise. The study also attempts to explore how these sub-factors affect digital engagement and is intended to understand the causal relationship between digital engagement, consumer-brand relationship, and purchase intentions. The study attempts to identify the effect of adjusting influencers authenticity, self-identity, and product fit set as moderating variables. The influencer who is the subject of this study was selected from a pool of active celebrity influencers by a panel of experts. Consumers who participated in the survey were asked to watch an influencer's YouTube, online presence, broadcasts, and SNS before completing the survey. A total of 528 consumers participated. The method of SEM with EQS 6 was used to verify the hypothesis in this study. The analysis results are as follows. Of the influencer value system sub-elements, attractiveness and expertise have a definitive impact on digital engagement. However, the sub-factor of influencer reputation has been found to have no significant effect on digital engagement. Digital engagement has been determined to have a positive impact on consumer-brand relationships, and consumer brand relationships have a positive impact on purchase intentions. Influencer authenticity and self-identity have moderating effects when attractiveness, reliability, and expertise affect the consumer-brand relationship.
Keywords:	Internet Influencer, Digital Engagement Consumer-Brand Relationship Purchase Intention
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	THE RELATIONSHIP BETWEEN E-GOVERNMENT EFFECTIVENESS AND E-GOVERNMENT USE: THE MEDIATING EFFECT OF ONLINE TRUST AND THE MODERATING EFFECT OF HABIT
Author:	SHADI AHMED KHATTAB, ISHAQ SHAAR, LINA AL ABBADI, AHMAD YOUSEF KALBOUNEH, WAEL BASHEER ALHYASAT
Abstract:	Governments in both rich and developing countries need to offer suitable e-government services to ensure citizens have trust and use them effectively and efficiently in today's digital and automated world. Users' trust and involvement in e-government have been negatively impacted by the current state of implementation, particularly in Jordan. Furthermore, to continue being responsive and effective, public administrations need to change the way they operate by utilizing more technology for communication and information. Client satisfaction and system trust will rise as a result of this. This article looks into the variables that influence the use of e-government. It further elaborates and investigates the mediating role of online trust in the relationship between e-government effectiveness and e-government use. It tests whether habit has any moderating effect on the relationship between e-government effectiveness and e-government. Data from 471 e-government users were used to support the research model, which was examined using a structural and measurement model using SmartPls 3.3.0. The findings show a direct relationship between e-government effectiveness and e-government use. Furthermore, the relationship between e-government use and efficacy is mediated by online trust. The results also show that habit moderates the relationship between the effectiveness and use of e-government. As a result, we can see the intention to utilize e-government again may be influenced by the practice of conducting business through the Internet. The implications for theory and practice in the area of e-government can drive public policy, and suggestions for future research are also highlighted.
Keywords:	E-government Effectiveness, E-government Use, Information Quality, System Quality, Service Quality, online Trust, habit.
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	CAAMP NET: AMP BASED CBAM ATTENTION NETWORK FOR CS RECOVERY
Author:	PAVITRA V , DR. SRILATHA INDIRA DUTT V.B.S
Abstract:	Deep learning based compressed sensing is a novel technology that has led to complete signal recovery with fewer measurements that are well below the required number of samples defined by Nyquist rate. It outperforms the existing non-deep learning-based methods by exploiting the inherent structures present in the signal or image of interest. There are iterative deep network translations of state of art recovery methods, as well as standard deep learning architectures in literature. The proposed method uses joint optimization of Compressed sensing as well as recovery using an Approximate Message Passing (AMP) based iterative method. This work also introduces use of attention gates along with AMP based recovery of compressively sensed image. Using attention helps the model to focus on high data areas and faster recovery. Experimental results show that joint optimization as well as using attention mechanism in the model out performs other Deep learning-based CS recovery methods in a range of measurement rates.
Keywords:	Approximate Message Passing, AMP, Attention, Compressed Sensing, Compressive ensing, CBAM Attention
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	ENERGY CONVERSION TECHNOLOGY TREND AND COMMERCIAZATION CASE USING ENERGY-ICT CONVERGENCE TECHNOLOGY
Author:	JONG-YUN KIM, YOUNG-SU KIM
Abstract:	This study explains the development trend of Energy-ICT convergence technology in Korea and major countries that is being promoted based on the policy of expanding advanced ICT infrastructure and knowledge information services, and the trend of expanding P2P power transactions, which is becoming a representative energy-ICT convergence platform. Based on this, we will present a plan to activate an Energy-ICT convergence business model based on distributed resources, a profit model through energy conversion, such as commercialization of domestic Energy-ICT convergence technology, and overseas commercialization cases(USA, Germany). The establishment of an environment that activates P2P power trading will undergo a very big change in the current paradigm of change in the electricity market. Therefore, it is necessary to have optimal solutions to many pending issues(power load/power market participant competitive structure/trading system, etc.). Furthermore, it is necessary to focus on building a foundation that can maximize the value of small-scale distributed resources by differentiating integrated control of distributed resources and services that can be traded in the electricity market through the advancement of core technologies.
Keywords:	Energy-ICT convergence, P2P power transactions, Business model, Distributed resources, Electricity market, Power load, Power market, Trading system, Integrated control
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	COMPARISON OF ACTIVITY-BASED COSTING AND TIME-DRIVEN ACTIVITY-BASED COSTING FOR PRINTED CIRCUIT BOARD ASSEMBLY PRODUCTION
Author:	NUR SYAFIKAH PINUEH, MOHD YAZID ABU, NURUL HAZIYANI ARIS, MUHAMMAD ARIEFFUDDIN MOHD JAMIL, EMELIA SARI
Abstract:	The number of resources employed determines product costs in traditional costing analysis (TCA). TCA is flawed because manufacturing overhead may be substantially larger than the basis of allocation, therefore a little change in resource volume causes a big change in overhead. Thus, activity-based costing (ABC) was created to solve TCAs cost allocation issues by analyzing overhead expenses and cost drivers more thoroughly. Many cost factors may be utilized to identify overhead causes and reduce overhead expenses. ABC was theoretically inappropriate to overlook spare capacity that may help forecasting. For that reason, time-driven activity-based costing (TDABC) has been studied because it maximizes capacity cost rate and time equations to create underused capacity information. This purpose of this study observes, analyses, and compares ABC and TDABC methods to determine which is more effective for production. Four steps comprise this studys approach. Phase 1 defines the problem, while phase 2 involves data collecting on location. Phase 3 implements ABC and TDABC methodologies and compares their costing analyses. The last phase, phase 4 concludes this research. As a result, this study shows ABC is transparent and can forecast unit product cost using cost driver rate. TDABC has objective cost driver determination, eliminates time-consuming processes, various cost drivers, and capacity utilization analysis for forecasting and planning. Thus, the comparative study met the third purpose of comparing ABC and TDABC costs sustainment utilizing numerous aspects. Finally, TDABC is the best practice because it provides better information than ABC in cost allocation, driver determination, action taken for an additional activity, cost consideration, informative, transparency, oversimplification of activities, and capacity forecast.
Keywords:	ABC, TDABC, PCBA, Cost Allocation, Cost Forecasting
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	DESIGN AND IMPLEMENTATION OF A CONTAINER ORCHESTRATION SYSTEM FOR DISTRIBUTED REINFORCEMENT LEARNING DATA ANALYSIS
Author:	SEOK JAE MOON, SEO YEON GU
Abstract:	Recently, reinforcement learning has shown excellent performance in solving complex data analysis problems in the real world, and many companies are actively introducing it. However, as the diversity and complexity of corporate business processes continues to increase, existing reinforcement learning approaches have limitations in solving complex problems. To cope with this, more sophisticated algorithms have been developed, but these algorithms require high computational resources. As a result, many enterprises seek to obtain the computing resources they need by leveraging distributed environments such as the cloud. However, as the diversity and complexity of enterprise business processes increase, the workload that cloud service providers must manage becomes more complex. Therefore, container orchestration mechanisms are becoming more complex and resource utilization is becoming more difficult. Therefore, in this study, we propose a container orchestration system for distributed reinforcement learning data analysis. This proposed system consists of User Interface, Task Processing Layer, and Infrastructure. In addition, through performance comparison experiments with existing centralized processing methods, that the proposed system is suitable for data analysis in a distributed environment.
Keywords:	Container Orchestration, Distributed Reinforcement Learning, Data Analysis
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	PERFORMANCE ANALYSIS OF NHPP-BASED SOFTWARE RELIABILITY MODEL WITH INVERSE-TYPE LIFE DISTRIBUTION PROPERTY
Author:	SEUNG KYU PARK
Abstract:	In this work, the performance of the NHPP-based software reliability model applying the Inverse-type distribution, which is widely utilized to various types of reliability life distributions, was newly identified. Accordingly, software failure time data was used to analyze reliability performance by predicting failures that may occur in the software operation, and the solution of parameters were estimated using maximum likelihood estimation. As a result, first, as a result of evaluating the criteria value (MSE and R^2) for efficient model selection, the efficiency of the Inverse-Exponential model was evaluated as the best. Second, as a result of analyzing the attributes data (m(t), λ(t), R ̂(τ)) that determine reliability performance, the Inverse-Exponential model was the most efficient. In conclusion, through various comparative analyses, the Inverse-Exponential model was found to have the best performance. Through the results of this study, the reliability performance of the Inverse-type life distribution for which there is no existing research data was newly analyzed, and basic design and test data necessary for an efficient software development process could also be presented. In the future, after exploring applicable statistical distributions for each software convergence industry, follow-up studies to find an optimal model will be needed.
Keywords:	Goel-Okumoto, Inverse-Exponential, Inverse-Rayleigh, Inverse-type, NHPP Model, Reliability Performance
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	A STUDY OF VIRTUAL REALITY CYBERSICKNESS AND CHANGES IN BINOCULAR VISUAL FUNCTION AND POSTURAL BALANCE
Author:	JAE-BEOM SON, JAE-MIN LEE, JEONG-LAE KIM, HYUN-SUNG LEEM
Abstract:	To analyze the causes of virtual reality (VR) cybersickness, which varies in severity from person to person after playing VR games using a head-mounted display (HMD) equipment, this study attempts to see how it affects binocular visual function and postural balance, and how it correlates with cybersickness. Before and after playing the VR game, the virtual reality sickness questionnaire (VRSQ) was used to objectively assess changes in cybersickness. Changes in binocular visual function were determined by measuring distance and near phoria, near point convergence (NPC), positive and negative fusional vergence at distance and near, and vergence facility. Finally, the stability index (SI) and fall risk index (FI) were confirmed using the TETRAX (balance ability test) to examine postural balance.Subjects were selected by controlling for variables such as age, corrected vision, and medical conditions that could affect the results. Significant changes in VRSQ, binocular visual function, and postural balance occurred after the VR game.
Keywords:	Virtual Reality(Vr), Head-Mounted Display, Cybersickness, Virtual Reality(Vr) Motion Sickness, Binocular Visual Function, Balance Test (Tetrax)
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	NEUMANN STACKED BILATERAL DEEP LEARNING BASED BIG SENTIMENT DATA ANALYTICS
Author:	M ANOOP, K SUTHA, UMA SHANKARI SRINIVASAN, M BALAMURUGAN, V ANITHA, J EMERSON RAJA
Abstract:	Sentiment analysis extracts information from several text sources like, blogs, reviews, news, and so on. The purpose of sentiment analysis on big data is to classify emotions or opinions into variegated sentiments. Conventional deep learning methods have been developed to classify the tweets. However, longer sentiment analysis time was considered. To address the issue, Neumann Mutual Informative and Stacked Bilateral Deep Learning (NMI-SBDL) for sentiment investigation is proposed to research products or services before making a purchase. First, through the tweets obtained from the Sentiment140 dataset, the Knowledge Sentimental Graph is constructed. Second, computationally-efficient dimensionality reduced tweets are generated by the Neumann Mutual Information-based Feature selection algorithm. Finally, the Stacked Bilateral LSTM-based model is utilized for classifying the tweet polarity. With this robust sentiment analysis is made by the Twitter Application Programming Interface (API) with higher accuracy and lesser computation time. Experimental assessment of the proposed NMI-SBDL and existing methods are carried out with different factors using Python libraries. The results of NMI-SBDL provided for improving the sentiment analysis accuracy, precision, recall and lesser time by 13%, 6%, 6%, and 23% than the existing approaches. The paper concludes with accurate and robust sentiment analysis for big data.
Keywords:	Big Data, Sentiment Analysis, Neumann Mutual Information, Feature Selection, Stacked Bilateral, Long Short-Term Memory
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	A STUDY ON STRESS REDUCTION USING INTELLIGENT EMOTIONAL INFORMATION SYSTEM
Author:	HYE-KYEONG KO
Abstract:	Stress is an integral part of modern life, and everyone experience if at some point. The stress experienced by teens has a wide range of mental and physical effects. This study examines an intelligent emotional information system that applies color therapy to measure the degree and cause of stress among adolescents with the goal of relieving it. Various types of emotional information can be gathered through emotional information design, which is important in understanding the end user as part of human-centered design, whereby a color information design method is proposed for each emotion using fuzzy neural networks.
Keywords:	Emotional Information System, Stress Reduction, Color Therapy, Fuzzy Neural Network
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	REGION-BASED FULLY DEEP CONVOLUTIONAL NEURAL NETWORKS ENHANCED WITH CARNIVOROUS PLANT ALGORITHM FOR PLANT DISEASE DETECTION AND CLASSIFICATION
Author:	E SARASWATHI , J FARITHA BANU
Abstract:	The pace of agricultural production is one of the most critical factors determining a nation's overall economic well-being. Diagnosing symptoms at an earlier stage can significantly reduce the spread of infectious diseases, increasing agricultural productivity. This research aims to advance agricultural technology and eventually boost agricultural productivity and financial outcomes by providing an effective approach for early detection and management of plant diseases. The automated detection of plant diseases is essential to plant monitoring since plant diseases are one of the most significant challenges faced in agriculture. This research proposes a method for recognizing and categorizing plant illnesses using region-based fully deep convolutional neural networks (RFDCNN). This method also utilizes Carnivorous Plant Algorithm (CPA) to enhance the average, mean precision of the results obtained using RFDCNN architecture and feature extraction. 230 plant disease leaf images like potato, Strawberry, pepper, peach, Squash, citrus, tomato, cherry, etc., from a farm field were gathered for this plant disease classification, along with a dataset of plant villages. According to the findings of the experiments, the RFDCNN-CPA deep learning framework that was suggested is able of accurately and efficiently classifying the many different varieties of plant leaves. The end results demonstrated that the RFDCNN-CPA obtained a higher accuracy rate of 97.92%, precision analysis of 95.82%, an f-score of 90.52%, and an overall execution time of 0.42ms.
Keywords:	Agriculture, Deep Learning, Plant Disease Detection and Classification, Region-based Fully Deep Convolutional Neural Networks, Computer Vision
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	AFFORDANCE-DRIVEN DESIGN FOR DIGITAL LEARNING
Author:	JIHWA NOH, YESON KIM , BYEONGSOO KIM
Abstract:	This investigation delves into the utilization of digital math learning platforms from an affordance design perspective, aiming to pinpoint crucial elements that significantly enrich learning experiences within the digital realm. Through an in-depth dissection of ten specific affordances categorized into Content, Pedagogy, and Functionality domains, this study meticulously evaluates how these components are manifested across six digital platforms widely adopted within South Korean public education. Employing a comprehensive analytical approach, the research uncovers notable deficiencies in the integration of these essential affordances, with a particular focus on the widespread presence of Nonlinearity contrasted against the stark underrepresentation of Adaptivity. These findings illuminate a prevalent design-practice gap, suggesting that the pedagogical potential of digital learning tools is not being fully harnessed to facilitate enhanced learning outcomes. The analysis further reveals how the integration or lack thereof of these affordances impacts the efficacy of digital learning environments, suggesting a critical need for a more holistic integration strategy. Contributing a structured analytical framework to the ongoing discourse on digital education, this study not only highlights current shortcomings but also paves the way for the development of next-generation learning technologies. These advancements are envisioned to support diverse educational strategies more effectively, promoting dynamic and adaptive learning environments that can better cater to the evolving needs and preferences of students. This research advocates for a paradigm shift in the design and development of digital learning platforms, aiming to fully exploit the transformative potential of digital education in fostering more engaging, personalized, and effective learning experiences.
Keywords:	Affordance, Digital learning, Learning platform, Mathematics Education
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	EMBEDDED PATTERN ANALYSIS OF PLANAR PHASED ARRAY ANTENNA FOR X BAND COMMUNICATION SYSTEMS
Author:	KALAIARASI. D, M.R.EBENEZAR JEBARANI
Abstract:	This study utilises the embedded element pattern to evaluate the Phased Array Antenna (PAA) in the context of the X-band communication system. All of them are linked to a finite array reference impedance. The central element of the embedded pattern is determined by a single element that is embedded within the finite array. The preparation of a Phased Array Antenna involves the use of a large number of radiating components that are connected to a phase shifter. This phase shifter is responsible for adjusting the phase of the radiating elements and producing a beam in the desired direction. PAA offers several advantages over a single radiating element, including increased power density, efficiency, directivity, and gain. This study examines the design and analysis of a rectangular phased array antenna commonly employed by military forces for satellite communications in the X-band frequency range.
Keywords:	Dipole Antenna, Directivity, Rectangular planar, Phased Array Antenna, Satellite Communication, X-band
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	EFFICIENT VIDEO COMPRESSION USING DEEP JOINT OPTIMIZATION METHOD WITH MOTION ESTIMATION AND INTER-FRAME PREDICTION
Author:	DR. M. CHANDRA SEKHAR, SAMPURNIMA PATTEM , B.SHRAVAN KUMAR DR. SUMAGNA PATNAIK
Abstract:	In the contemporary era, there is unprecedented increase in multimedia content, especially videos, leading to consumption of more bandwidth when transmitted. Video compression is the technique that leverages performance of video transmission as it reduces original size of the video. Though the conventional video compression methods have classical architecture to encode motion and residual information efficiently, it lacks the ability to have non-linear representation of data. In this paper, we proposed a framework named Artificial Intelligence (AI) enabled Video Compression Framework (AIVCF) which exploits the traditional classical architecture and combines it with a deep learning model for non-linear data representation. This framework has ability to have joint optimization of underlying components. Convolutional Neural Network (CNN) is used to reconstruct current frames by getting motion information through a process known as optical flow estimation. The information of given video is compressed using deep learning models in auto-encoder fashion. The framework strikes balance between quality and compression ability. An algorithm named Deep Joint Optimization for Video Compression (DJO-VC) is proposed to realize the AIVCF. The proposed framework is evaluated with empirical study. The experimental results, in terms of PSNR and SSIM revealed that the proposed framework outperforms existing models such as H.264.
Keywords:	Video Compression, Deep Learning, Convolutional Neural Network, Artificial Intelligence Enabled Video Compression Framework
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	A GLOBAL REVIEW OF CLOUD COMPUTING ADOPTION IN SMALL AND MEDIUM ENTERPRISE
Author:	RUSLAINI , DARMAWAN NAPITUPULU
Abstract:	Cloud computing is an emerging computing architecture that has gained favor among government agencies and businesses that need large-scale, low-cost computing. It allows services to be delivered in a scalable manner using machines in massive data centers. Cloud computing technology is becoming more and more significant as small and medium enterprises (SMEs) go through a digital transformation. In addition to enhancing accessibility, scalability, efficiency, and data security, it might help SMEs save money. Though several studies have been conducted on the subject, there is still a paucity of bibliometric research on the use of cloud computing in the SMEs sector. Bibliometric analyses provide a crucial role in understanding the evolution of academic research in particular fields. Consequently, a detailed examination of the trend and acceptance of cloud computing in the public sector is needed. To achieve this, bibliometric performance analysis and scientific mapping methodologies were used to analyze most publications and citations on cloud computing adoption in SMEs in this study. Finding the most productive authors, organizations, countries, and publishing sources was part of this. In addition, a clustering analysis was carried out to identify the primary research theme about SMEs' use of cloud computing during a ten-year period. With the use of the statistical programs Vosviewer and Biblioshiny, the dataset is shown and conclusions are reached while accounting for 246 articles that were extracted from the Scopus database and produced data between 2010 and 2024. The results of this study have added significantly to the body of literature and current knowledge by deepening our understanding of how cloud computing is used in SMEs.
Keywords:	Cloud Computing, SME, Adoption, Bibliometric Analysis, Review
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	STRENGTHENING PAYMENT CARD DATA SECURITY: A STUDY ON COMPLIANCE ENHANCEMENT AND RISK MITIGATION THROUGH MFA IMPLEMENTATION UNDER PCI DSS 4.0
Author:	PIETERS NICHOLAS PARADONGAN TAMBUNAN , NILO LEGOWO , DENNIS RYDARTO TAMBUNAN
Abstract:	The increasing use of electronic payments and the growing reliance on payment card transactions have underscored the importance of robust security measures to protect payment card data. The Payment Card Industry Data Security Standard (PCI DSS) has long been recognized as a crucial framework for ensuring security and compliance in handling payment card data. Amidst evolving cyber threats, the adoption of Multi-Factor Authentication (MFA) has emerged as a critical strategy to enhance the security of payment card data, improve compliance with PCI DSS 4.0, and mitigate associated risks. This study involves payment gateway organizations subject to PCI DSS 4.0 compliance requirements. Qualitative data confirms the effectiveness of MFA in thwarting cyber threats and enhancing overall payment card data security. In an era marked by evolving cyber threats, this research emphasizes the importance of implementing MFA to bolster payment card data security, achieve compliance with PCI DSS 4.0, and mitigate risks. The findings of this study offer actionable insights for organizations seeking to strengthen their payment card data security measures and align with regulatory standards.
Keywords:	Payment Card Data Security, PCI DSS 4.0, Multi-Factor Authentication (MFA), Compliance Enhancement, Risk Mitigation.
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	ADVANCED HYBRID PREDICTION MODEL: OPTIMIZING LIGHTGBM, XGBOOST, LASSO REGRESSION, AND RANDOM FOREST WITH BAYESIAN OPTIMIZATION
Author:	MR. SANJAY KUMAR , DR. MEENAKSHI SRIVASTAVA , DR.VIJAY PRAKASH
Abstract:	This paper proposes an advanced hybrid prediction model that combines the strengths of LightGBM, XGBoost, Lasso Regression, and Random Forest. To optimize the hyperparameters of these diverse models, we leverage Bayesian Optimization, a powerful technique for efficient hyperparameter search. The proposed model integrates predictions from optimized individual models, potentially leading to improved accuracy and robustness. Our experimental evaluation demonstrates the effectiveness of the proposed hybrid model compared to baseline models. This study compares the performance of various regression models, including Random Forest, Lasso Regressor, XGBoost Regressor, and LightGBM, against a proposed hybrid model. Evaluation metrics such as Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), R2 Score, Explained Variance Score (EVS), Mean Absolute Percentage Error (MAPE), and Mean Percentage Error (MPE) are analyzed. The proposed hybrid model demonstrates superior performance across all metrics, with observed values of 1.622813174 for MSE, 1.273896846 for RMSE, 0.652113986 for MAE, 0.99996681 for R2 Score, 0.999966815 for EVS, 0.177679435 for MAPE, and -0.001810521 for MPE. These results show the potential of the hybrid model for accurate prediction in regression tasks. This research contributes to the field of advanced prediction modeling by offering a novel hybrid approach that leverages Bayesian Optimization for improved performance and interpretability. In future work, researchers plan to explore additional machine learning algorithms and optimization techniques to further enhance the performance of the hybrid model. The hybrid prediction model developed in this study holds great promise for advancing predictive analytics and decision support systems in diverse application domains.
Keywords:	LightGBM, XGBoost, Lasso Regression, Random Forest, Bayesian Optimization, Hybrid prediction model, Interpretability.
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	COLOR IMAGE ENCRYPTION BASED ON ARNOLD CAT MAP- ELLIPTIC CURVE KEY AND A HILL CIPHER
Author:	DESAM VAMSI , PRADEEP REDDY CH
Abstract:	The role of Image encryption in communication over the internet is gaining a lot of interest in recent days with the increased use of the internet. Transfer of crucial images over the internet is prone to attack and stealing when shared on an unsecured medium. To avoid these attacks and theft of image data, encryption is the best process to be used. In this paper, a novel approach of sub- image shuffled Arnold cat followed by an elliptic curve is used for encryption and decryption. The RGB planes of the color image are extracted, and each plane is divided into four sub-images that undergo the Arnold cat transformation that effectively scrambles and shuffles the image values, and four sub-images are combined back. The Arnold transformed RGB planes are combined into a single image that is further processed with an Elliptic curve key to compute a self-invertible key matrix and perform Hill cipher operation for encryption and decryption. The proposed method generates good quality cipher images with an entropy value of 7.9, low correlation, and can resist statistical and differential attacks.
Keywords:	Encryption, Arnold Cat, Elliptic Curve, Self-Invertible Key Matrix, Hill Cipher, RGB Planes
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	AI ENABLED SYSTEM WITH REAL TIME MONITORING OF PUBLIC SURVEILLANCE VIDEOS FOR ABNORMALITY DETECTION AND NOTIFICATION
Author:	PRATHAP ABBAREDDY, DR. SK. YAKOOB, RAMU KUCHIPUDI, BHUJANGA REDDY BHAVANAM
Abstract:	Public surveillance videos are increasingly playing key role in identification of certain incidents and people who misbehave or perform illegal activities. Monitoring surveillance videos manually to detect abnormalities is time consuming and it may lead to delay in getting required information. With the usage of Artificial Intelligence (AI) video analytics in real time can help in acquiring such information on time so as to make well informed decisions. Particularly deep learning is great help in learning from incidents and detect anomalous behaviours. In this study, we suggested an autonomous system for anomaly detection from surveillance films, based on deep learning. For anomaly detection, an improved Convolutional Neural Network (CNN) model is employed. We presented a method that utilizes the upgraded CNN model for its functionality, called Learning based Video Anomaly Detection (LbVAD). To lower the prediction process's error rate, a loss function is defined. For our empirical investigation, we gathered data from many benchmark datasets, including UMN, UCSD, Ped1, and Ped2. The suggested approach works better than the current models, according to the results of our experiments.
Keywords:	Machine Learning, Deep Learning, Artificial Intelligence, Video Abnormality Detection
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	A HYBRID MALWARE DETECTION FRAMEWORK WITH DRIFT ADAPTATION FOR TIMESTAMPED DATA
Author:	HARINADH VARIKUTI , VALLI KUMARI VATSAVAYI
Abstract:	As technology is growing rapidly, new malware variants are evolving. Attackers are generating new malware patterns using code obfuscation and mutation techniques to escape from anti-malware engines and traditional models. Models built over historical malware data cannot perform well in detecting or classifying dynamic or real-time malware data. The Evolution of new malware patterns may lead to concept drift. Static models built over historical data trained from scratch whenever new data arrives which is time-consuming, so models with adaptive nature are used. Adaptive machine learning models are used to identify the drift in data and perform appropriate changes dynamically to the models over time. In this paper, a hybrid model is proposed which is the combination of Leveraging bagging and AdaBoost methods for incremental ensemble learning. Dynamic weights are assigned to the models used in the ensemble learning to increase the adaptivity. The use of boosting and bagging methods adds error adoption and diversity to the model respectively. Experiments also show the proposed model outperforms all the state-of-the-art drift adaption methods such as ADWIN Bagging, ADWIN Boosting, SRP Classifier, etc.
Keywords:	Concept drift, Machine learning, Malware, drift detection, Bagging, Boosting
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	FINANCIAL REVOLUTION: INNOVATION POWERED BY FINTECH AND ARTIFICIAL INTELLIGENCE
Author:	HASSAN CHIKRI , MANAR KASSOU
Abstract:	The article explores the dynamic merger between Fintech (financial technology) and Artificial Intelligence (AI), marking the start of a global financial revolution. The partnership between Fintech and AI is becoming the driving force behind significant revolution in the finance industry, reshaping our approach, use, and comprehension of financial matters in an environment that is always evolving. An in-depth analysis of these recent advancements brings attention to the untapped possibilities and difficulties that arise from this merging of technologies. The development of Fintech, which occurs at the intersection of finance and technology, centers on two essential elements: providing client service and ensuring transaction security. Fintech aims to enhance customer experience by creating a more customized and user-friendly financial environment. Intuitive applications, automated financial advisers, and easily available services remove conventional obstacles, enabling quick and transparent transactions. Fintech prioritizes transaction security with the use of sophisticated processes, including biometrics and two-factor authentication, bolstered by technology like machine learning to identify and prevent fraud. Regtech solutions are used to guarantee adherence to regulatory requirements, streamlining the administration of complex financial rules. The paper emphasizes the significance of agility and innovation in the Fintech industry, specifically stressing AI as a crucial catalyst for innovation. Artificial intelligence (AI) has a significant influence on several industries, ranging from banking to medical research, via the automation of intricate procedures, enhancement of service customisation, and reinforcement of transaction security. The amalgamation of Fintech with AI has substantial advantages, fundamentally transforming the manner in which financial services are provided, used, and controlled. Artificial intelligence facilitates thorough examination of financial data, identification of fraudulent activities, and automation of processes, hence facilitating prompt and well-informed decision-making. Nevertheless, the paper also emphasizes the difficulties associated with data security, ethical ramifications, and regulatory apprehensions in this constantly evolving digital landscape. Artificial intelligence (AI) is playing a pivotal role in the financial revolution, introducing substantial modifications that redefine the norms of the financial sector. Fintech, driven by artificial intelligence, offers a smooth, easily accessible, and highly secure experience, which is reshaping the future of finance in the contemporary digital environment.
Keywords:	Fintech, Artificial Intelligence, Financial Revolution, Financial Innovation, Blockchain Technology
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	EVALUATING THE PERFORMANCE OF XGBOOST AND GRADIENT BOOST MODELS WITH FEATURE EXTRACTION IN FMCG DEMAND FORECASTING: A FEATURE-ENRICHED COMPARATIVE STUDY
Author:	MURARI THEJOVATHI , DR M.V.P. CHANDRA SEKHARA RAO
Abstract:	In this paper We are proposing the inclusion of Gradient Boost, another ensemble technique, to broaden the scope and potentially improve forecasting accuracy. This research looks at how XGBoost and Gradient Boost, two powerful ensemble learning methods, can be used to predict demand in the FMCG sector. The suggested method also includes advanced feature extraction techniques to make the model work better. The current method uses XGBoost, a well-known and effective gradient-boosting technique that is fast and easy to scale. The suggested method includes gradient boost, which is another ensemble technique, as well as feature extraction techniques that help find and use the dataset's most important information. The research aims to compare the performance of XGBoost and Gradient Boost models in the context of demand forecasting for Fast-Moving Consumer Goods (FMCG) data. Additionally, the study incorporates feature extraction methods to enhance the models' predictive capabilities. We test both models thoroughly using FMCG data to see how well they work in terms of accuracy, reliability, and how quickly they can be run. To find the factors that have the most influence on demand prediction, feature extraction techniques like Principal Component Analysis (PCA) and Recursive Feature Elimination (RFE) are used. The study's results tell us a lot about how well the XGBoost and Gradient Boost models work for predicting demand in the FMCG sector. Using feature extraction methods is also meant to find hidden patterns in the data, which will help supply chain professionals in the FMCG business make more accurate predictions and better decisions. Researchers can use the results of this study to help them choose the best method for their own demand forecasting needs. This will improve operational efficiency and cut costs in the FMCG supply chain.
Keywords:	GaradientBoost, XGBoost, FMCG Sector, Feature Extraction, Demand forecasting
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	ENHANCED RICE PLANT DISEASE IDENTIFICATION: A HYBRID APPROACH OF TRANSFER LEARNING, SVM AND PCA
Author:	SURENDER MOGILICHARLA , UPENDRA KUMAR MUMMADI
Abstract:	The accurate and timely identification of diseases and pests impacting rice cultivation is crucial for farmers, allowing for swift intervention and thereby minimizing economic losses. Recent progress in convolutional neural networks (CNNs) has significantly boosted the accuracy of image classification, yet their resource-intensive nature, demanding significant memory and processing power, underscores the necessity of leveraging pre trained models. Additionally, the paper introduces an ensemble approach combining deep learning with traditional machine learning methods, further enhancing the effectiveness of disease and pest detection in agricultural settings. The paper presents a novel approach leveraging state-of-the-art large-scale architecture, ResNet-50, to propose two distinct models: Model-1 integrates ResNet-50 with SVM, while Model-2 incorporates ResNet-50 with PCA and SVM to effectively detect and identify rice diseases and pests. Through experimentation on authentic and real datasets, the paper demonstrates the effectiveness of these models. Additionally, recognizing the constraints of large-scale architectures, especially concerning their compatibility with mobile or embedded devices due to processing power and memory limitations, the paper introduces and evaluates two proposed models. Notably, Model-2 surpasses Model-1, achieving a superior accuracy of 93.7% compared to 91.6% in Model-1. Moreover, Model-2 significantly reduces the feature set size by 100% compared to Model-1 through dimensionality reduction using PCA.
Keywords:	Deep Learning, Ensemble Methods, Rice Plant Disease Detection, Pretrained Models, Transfer Learning
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	EA HEART DISEASE DIAGNOSIS SYSTEM EMPLOYING RESIDUAL CONVOLUTIONAL NEURAL NETWORKS WITH ADAPTIVE CROSS-LAYER STACKED ARCHITECTUREA
Author:	SLAKSHMI ANUSHA KOTHAMASU, MAHANKALI SARITHA, B SWATHI, KOORAGAYALA SUKEERTHI
Abstract:	When all diseases are considered, heart disease is the leading cause of mortality worldwide. Heart disease is a significant risk. Whenever the blood and oxygen-supplying arteries to the heart become completely blocked or constricted, a cardiac issue results. Heart disease is one of the primary causes of death. In a short period of time, there has been a substantial increase in mortality. Accurate and secure diagnosis should be given priority in healthcare since a misdiagnosis of heart illness may result in mortality. If the prognosis is true, although cardiovascular disease may be avoided, if the prediction is incorrect, it may be dangerous. As a result, the team has created an ACLS-RCNN based ICSOA technique-based Intelligent heart disease prediction model. In the beginning, the patient data is acquired and pre-processed. Handling missing value handling, the three primary components of data pre-processing include scaling with IQR-RS, managing imbalance data using ROS, and pre-processing the data itself. Features are extracted using Kernel-based Linear Discriminate Analysis after the pre-processed data. Then, to eliminate pertinent components, Utilizing hybrid optimization, also known as (CI-AO+GI-SMO), valuable data is selected. Using recommended Cross-Layer Stacked Residual Convolutional Neural Networks that are Adaptive, an illness will be predicted. The prediction technique is improved by using the Improved Cuttlefish-Swarm Optimization Algorithm (ICSOA). Using Python simulation, the recall, accuracy, f1-score and precision of the suggested system are compared to those of other well-known methods. With regard to clinical data and diagnoses, our technology is capable of making exact predictions.
Keywords:	Improved Cuttlefish-Swarm Optimization, Convolutional Neural Network, Heart Disease, Correlation-based feature selection, Kernel-based Linear Discriminant Analysis
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text

Title:	SENTIMENT-BASED RECOMMENDATION FOR ONLINE SHOPPING
Author:	MAFAS RAHEEM, NIRASE FATHIMA ABUBACKER, DEVINA WIYANI
Abstract:	In E-commerce, a widely used strategy to improve the customer shopping experience and address information overload is the implementation of recommendation systems. Many E-commerce platforms have their proprietary recommendation algorithms, with content-based filtering being a commonly employed approach. This algorithm provides non-personalized suggestions to users based on the similarity of content. The shopping experience involves the decision-making process, with a significant focus on information search. In online shopping, customers heavily rely on information search to gain in-depth insights into products, as physical interaction is not possible. Customer reviews play a crucial role in this information search, offering shoppers the opportunity to learn from the experiences of previous customers. Despite the importance of customer reviews, existing recommendation solutions often overlook this aspect in their product recommendations. To address this gap, sentiment analysis, a natural language processing task frequently used in reviews, is employed to classify, and quantify text based on its polarity. This research introduces a recommendation pipeline that combines content-based filtering, utilizing cosine similarity calculations, and sentiment analysis, utilizing a pre-trained RoBERTa language model. The focus is on quantifying customer reviews from an E-commerce platform in Malaysia. The goal of this research is to develop an embedded system that recommends products to users based not only on their similarity but also on high ratings from various E-commerce sites.
Keywords:	Sentiment Analysis, Content-Based Recommendation, Sentiment-Based Ranking, Product Review Analysis
Source:	Journal of Theoretical and Applied Information Technology 15^th May 2024 -- Vol. 102. No. 9-- 2024
Full Text