		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
April 2025 | Vol. 103 No.8

Title:	ENHANCING SQL INJECTION (SQLI) MITIGATION BY REMOVING MALICIOUS SQL PARAMETER VALUES USING LONG SHORT-TERM MEMORY (LSTM) NEURAL NETWORKS
Author:	NWABUDIKE AUGUSTINE, ABU BAKAR MD. SULTAN, MOHD HAFEEZ OSMAN, KHAIRONI YATIM SHARIF
Abstract:	Web applications are increasingly targeted by cyber attacks. With SQL injection being a significant vulnerability, it caused an estimated USD4 billion global economic loss in 2022. Research has been conducted to explore methods to mitigate the impact of these attacks either by detecting them immediately or preventing them in their tracks. However, conventional approaches such as rule-based or signature-based systems have limitations as they cannot adapt to new or obscure attack patterns. This study explores the potentials of using Long Short-Term Memory (LSTM) neural network-based architectures for sanitization of SQL parameter values from malicious characters and shows that the result demonstrates that LSTM is the top performer, consistently achieving near-perfect accuracy, precision, recall, and F1-scores of 99.66% effectively. In a nutshell, this work has made a very strong and scalable contribution to web application security problems and demonstrates the possibility of SQLi mitigation using LSTM networks. By addressing such a critical gap in the literature, this study is a significant progress in the field of cyber security and robustness against obfuscation of attacks.
Keywords:	SQL Injection, Long Short-Term Memory, Web Application Security, Machine Learning, Cyber Attack
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	USING BIG DATA TO DEVELOP DIGITAL MARKETING STRATEGIES: A CASE STUDY
Author:	YULIIA SOKOLOVA, OLHA KATUNINA , NADIIA PYSARENKO , INNA KLIMOVA , OLEH KOVALCHUK
Abstract:	The power of Big Data to deliver tailored, hyper-focused content to audiences, segment audiences into factions, forecast trends, and fine-tune marketing and advertising spending is hard at work in today’s competitive digital landscape – and, for many businesses, challenging to see. So, this study focuses on exploring the role of Big Data in the technique development of digital marketing strategies, with Amazon as a case study over the period 2018–2023. The goal is to study how data-driven strategies influence the most crucial marketing outcome metrics, including customer engagement, conversion rates, return on investment (ROI), and advertising efficiency. The study uses a mixed-method approach to combine Amazon's qualitative content analysis method and quantitative analysis via regression models and Difference-in-Differences (DiD) estimation. Results suggest that personalisation, segmentation, and real-time advertising optimisation significantly improve performance – post-intervention campaigns increase conversion rates by 3.1% and reduce customer acquisition costs by $10. Further, regression analysis reinforces the positive effect of Big Data strategies, and more profoundly, advertising optimisation has the most significant impact on ROI. Based on the findings in this study, the study recommends that policy be based on continuous loops of feedback and predictive analytics to facilitate adaptive, efficient, and targeted marketing efforts. However, it will also be critical to establish clear data governance frameworks and ethical data practices to maintain consumer trust and compliance with privacy regulations.
Keywords:	Content Personalisation, Market Trends Forecasting, Advertising Campaigns Optimisation, Audience Segmentation.
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	LEVEL OF GPT CHAT USE AMONG STUDENTS AND FACTORS AFFECTING ITS USE
Author:	BERTON ATALLAH BRAHMACARI PARIKESIT, Drs. TUGA MAURITSIUS
Abstract:	One of the preferred AI language models that can be used in natural language interaction with consumers is Open AI's ChatGPT. Based on a quantitative research methodology with a questionnaire survey method, this study examines the factors that influence users' acceptance and use of ChatGPT using the theory of acceptance model (TAM). Respondents in this study were 100 ChatGPT users. Acceptance and use of ChatGPT through the theory of acceptance and use (TAM). The findings are six factors, there are five factors accepted, but one factor is rejected, namely the perceived interactivity factor User participation with ChatGPT is explained by perceived interactivity and shows that in the future people will use chat gpt to help them in terms of learning.
Keywords:	Chat GPT, TAM, PLS-SEM, Education,
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	ZERO TRUST SECURITY FOR CARD-NOT-PRESENT TRANSACTIONS: EXTENDING EMV-LIKE CONTINUOUS AUTHENTICATION AND ADAPTIVE RISK VALIDATION ACROSS PAYMENT NETWORKS
Author:	MAROUANE AIT SAID , ABDELMAJID HAJAMI , AYOUB KRARI
Abstract:	Card-Not-Present (CNP) fraud remains a critical challenge [1][2][3] in digital payments, exploiting gaps between merchants, acquirers, and issuers within trusted payment networks. While EMV technology ensures dynamic authentication for Card-Present (CP) transactions [4][5], CNP transactions lack equivalent protection [6], often bypassing real-time risk assessment. This pa-per introduces a Zero Trust security model for CNP transactions, extending EMV-like continuous authentication and adaptive risk validation across payment stakeholders without modifying the ISO8583 messaging standard. By leveraging AI-driven risk scoring, behavioral biometrics, device finger-printing, and multi-factor authentication (MFA), the model ensures continuous verification from initiation to authorization. Risk scores dynamically evolve across the payment chain, enabling real-time decision-making. Experimental results demonstrate a 92.1% fraud detection accuracy, a 36% reduction in false positives, and real-time processing within 310 milliseconds per transaction. This approach bridges the security gap in CNP transactions, aligning with PCI-DSS, PSD2, and EMVCo standards while preserving user experience. By extending Zero Trust principles across the payment network, this work establishes a scalable and resilient framework for securing digital transactions.
Keywords:	Zero Trust Security, Card-Not-Present Fraud, EMV-Like Authentication, ISO8583, Continuous Authentication, Adaptive Risk Validation, AI-Driven Fraud Detection.
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	A STUDY OF ONLINE VIDEO CONSUMPTION BEHAVIOR THROUGH SOCIAL MEDIA AMONG GENERATION Z IN INDONESIA
Author:	BRYAN HENRY , VIANY UTAMI TJHIN
Abstract:	The advancement of technology has also led to the increasing prominence of social media. The exposure to social media, particularly among Gen Z, has raised concerns about the potential for online video addiction on these platforms. This study aims to analyze the factors influencing social media use and its potential impact on online video addiction among Gen Z. The research employs a quantitative approach using the Structural Equation Modeling (SEM) method. Data collection was conducted through a questionnaire completed by 400 respondents. The findings indicate that factors such as internet literacy and loneliness have a significant positive effect, whereas FOMO does not have a significant positive impact on social media use and the potential for online video addiction. Additionally, sensation seeking does not have a significant positive moderating effect on social media use and, therefore, does not influence the potential for online video addiction.
Keywords:	Internet Literacy, Loneliness, FOMO, Social Media Use, Online Video Addiction, Sensation Seeking
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	AI-DRIVEN FRAUD DETECTION AND SECURITY SOLUTIONS: ENHANCING ACCURACY IN FINANCIAL SYSTEMS
Author:	JANJHYAM VENKATA NAGA RAMESH , DR SUDHANSU SEKHAR NANDA , VAMSI KRISHNA CHIDIPOTHU , VEMULA JASMINE SOWMYA , AMIT VERMA , TAOUFIK SAIDANI
Abstract:	Financial institutions worldwide are facing a major challenge in identifying unauthorized transactions. Financial organizations need sophisticated fraud detection systems to safeguard their financial stability and maintain customer confidence; nevertheless, the design of such systems may be hindered by many complexities. The infrequency of illegal transactions and the disproportion in several transaction datasets, namely the few occurrences of fraudulent transactions relative to legitimate ones, exemplify these characteristics. The dataset disparity may undermine the accuracy and effectiveness of a fraudulent activity detection program. Disseminating customer information to provide a more effective centralized approach is impractical, since each bank is required to comply with data protection standards. To ensure the user experience remains unaltered, the fraud detection solution must be both accessible and easily observable. As a result, this study presents an innovative approach in tackling these issues by combining Federated Learning (FL) with Explainable AI (XAI). FL financial institutions work together to train a model that can identify fraudulent activities without sharing client information directly. This protects the privacy and security of client information. The addition of XAI makes it definite that human specialists can understand and analyse the model's results, which gives the system more reliability and transparency. The FL-based fraudulent detection system regularly shows high-performance measures, as shown by experiments employing real transaction datasets. This study demonstrates that federated learning might serve as an effective and confidential instrument in combating fraudulent activity.
Keywords:	Federated Learning, Explainable AI, Fraud Detection Systems, Data Privacy, Client Information
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	THE ROLE OF DIGITAL TWIN IN OPTIMIZING SMART LAUNDRY SERVICES: ANALYZING CUSTOMER ACCEPTANCE AND LOYALTY THROUGH TAM AND IS SUCCESS MODEL
Author:	SHIFA REGITA , VIANY UTAMI TJHIN
Abstract:	Traditional laundry services often struggle with inefficiencies, high operational costs, and a lack of real-time monitoring. To address these challenges, this study examines the role of Digital Twin (DT) technology in optimizing smart laundry services. DT enables real-time data synchronization, predictive maintenance, and enhanced decision-making by creating a virtual replica of physical assets. This research adopts the Technology Acceptance Model (TAM) and the IS Success Model to analyze customer acceptance and loyalty toward DT-based laundry services. The findings confirm that Information Quality, System Quality, and Service Quality significantly influence Perceived Usefulness and Perceived Ease of Use, which in turn impact Customer Satisfaction and Customer Loyalty. The study also highlights that accurate information, a stable system, and high-quality service contribute to a positive user experience. By integrating DT, smart laundry services can improve operational efficiency, reduce downtime, and enhance customer trust. However, challenges such as limited market awareness and the adaptation of DT in service-based businesses remain. The results reinforce that DT is a promising innovation for modernizing laundry operations, making them more reliable, efficient, and customer-centric.
Keywords:	Digital Twin, Smart Laundry, Customer Satisfaction, Customer Loyalty TAM, IS Success Model
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	GENERATING OPTIMAL TEST CASES USING ELITIST GENETIC ALGORITHM
Author:	ASHOK KUMAR BANDLA, C S PAVAN KUMAR, DR K KOTESWARA RAO, DR O. RAMA DEVI, DR KALAIVANI K, KOLACHANA SWETHA, DR. CHANDANAPALLI SURESH BABU
Abstract:	In order to reduce the number of test cases that do not significantly improve the mean of test coverage or where the test cases are unable to isolate mistakes, this research study examines the use and efficacy of the Elitist genetic algorithm. In this work, a genetic algorithm is used to help minimize or optimize the test cases. The algorithm creates the initial population at random, determines the fitness value using coverage metrics, and then uses genetic operations such as selection, cross-over, and mutation to select the offspring in successive generations. Specific genetic modeling processes may differ from standard genetic algorithms depending on the task. This generation process is performed until the fitness values remain unchanged for two successive generations. Convergence or a reduced test case is reached when the data generation remains unchanged for two iterations. The findings of study reveal that, genetic algorithms can greatly reduce the amount of the test cases
Keywords:	Elisist GA, Test Case, Optimal, NP Complete, Minimize
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	AN OPTIMIZED DEEP LEARNING MODEL FOR EARLY DETECTION OF RHEUMATOID ARTHRITIS USING KNEE X-RAYS
Author:	SALONI FATHIMA , G. SHANKAR LINGAM
Abstract:	Rheumatoid arthritis (RA) is a chronic autoimmune condition that requires early diagnosis to prevent irreversible joint damage. While deep learning models have shown promise in automated RA detection, existing approaches suffer from data imbalance, suboptimal feature extraction, and poor generalizability, limiting their clinical applicability. This study proposes an optimized Convolutional Neural Network (CNN) model, incorporating image augmentation and class balancing techniques to improve RA detection accuracy from knee X-ray images. Unlike previous studies, which often overlook the impact of data augmentation on model performance, our work demonstrates its effectiveness in addressing data imbalance and enhancing model robustness. We trained and validated our model using the Kaggle knee X-ray dataset, applying image augmentation to expand training samples and oversampling to balance class distributions. The CNN was optimized through rigorous hyperparameter tuning. Our optimized CNN model achieved a high accuracy of 94%, outperforming baseline deep learning models. Data augmentation and oversampling significantly improved model performance, proving their effectiveness in medical imaging tasks. Our findings establish a novel deep learning framework for RA detection, demonstrating the importance of data augmentation and optimization in improving diagnostic accuracy. This work contributes to the growing field of AI in medical imaging by offering a scalable and interpretable solution for automated RA detection
Keywords:	Rheumatoid Arthritis, knee X-ray images, statistical augmentation, deep learning, CNN.
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	OPTIMIZING BUOYANT WIRELESS SENSOR NETWORKS: EVALUATING ALBATROSS ADAPTIVE ACOUSTIC ROUTING PROTOCOL FOR ENHANCED PERFORMANCE AND RELIABILITY
Author:	DIVYA JOSE J , Dr. D. VIMAL KUMAR
Abstract:	The proposed Albatross Adaptive Acoustic Routing Optimization (AARO) protocol represents a breakthrough in IT-driven solutions for underwater wireless sensor networks. By leveraging adaptive routing mechanisms and bio-inspired optimization, AARO addresses critical challenges in underwater communication, including high energy consumption, network instability, and packet loss. This research contributes to IT by integrating intelligent routing decisions based on real-time environmental data, enhancing network longevity, and improving data transmission reliability. The findings demonstrate that AARO significantly outperforms existing protocols, offering a scalable and energy-efficient solution for buoyant WSNs. The research aligns with advancements in IT, enabling efficient data acquisition in marine exploration, disaster management, and environmental monitoring, reinforcing the role of IT in solving real-world challenges.
Keywords:	Buoyant WSN - Adaptive Acoustic Routing Protocol - Underwater Communication - Routing - Albatross Optimization.
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	A NOVEL METHOD FOR FLEXIBLE CAPACITY MULTI-IMAGE HIDING
Author:	HESHAM F.ABDELRAZIK , AHMED S.ELSAYED , SALLY S.ISMAIL , ABEER M.MAHMOUD
Abstract:	The noteworthiness of multiple image-hiding has increased in the current digital era. This owes to the growing demand for robust data protection and secure communication. The success of the multiple image hiding algorithm depends on concealing multiple secret images in one cover image and perfectly recovering all secret images at the receiving end. Unfortunately, the two primary challenges with current multi-image hiding methods are preserving high capacity while reducing visual distortions and guaranteeing precise recovery of every hidden image. The main challenge is to keep the hidden image visually obscured while concealing as much information as possible from a single cover image. In this paper, a novel framework for concealing multiple images using an invertible hiding neural network and image super-resolution is proposed. Extensive experiments on DIV2K, ImageNet, and COCO demonstrated that the proposed framework obtained significant results in terms of both imperceptibility and recovery accuracy of the secret image. Achieving an average of 27.16 dB, 26.65 dB, and 32.13 dB of PSNR and 0.810, 0.783, and 0.885 of SSIM for hiding four secret images with upscaling factor 2, respectively. The proposed framework successfully concealed and revealed up to 64 secret images. These results confirm the effectiveness of the proposed method in achieving both high invisibility and enabling multiple image concealment with a large capacity.
Keywords:	Image hiding, Image rescaling, Invertible hidden neural network, Image super-resolution, Deep learning.
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	DISEASE RISK PREDICTION USING ELECTRONIC HEALTH RECORD DATA BASED ON FLYING SQUIRREL SEARCH OPTIMIZATION WITH BIDIRECTIONAL – LONG SHORT TERM MEMORY
Author:	PRASANTHI YAVANAMANDHA , D. S. RAO
Abstract:	Nowadays, the attention towards effective disease risk prediction has increased, due to its importance in monitoring the future health status of patients. This prediction helps to provide the right treatment for patients to prevent severe stages of diseases. However, the existing risk prediction models based on Machine Learning (ML) have limitations in learning temporal information from the Electronic Health Record (EHR) data. To overcome this, a Flying Squirrel Search Optimization (FSSO) algorithm for feature selection and Bi-directional Long Short Term Memory (Bi-LSTM) is proposed to enhance the accurate disease risk prediction using EHR data. The proposed FSSO based feature selection method efficiently reduces the dimensionality of EHR data and helps to identify the most predictive features for disease risk accurately. By utilizing the bidirectional layers, Bi-LSTM model learns the dependencies in both past and future directions, which makes the model suitable for capturing comprehensive temporal patterns that influence disease risk. Initially, the EHR data is acquired and preprocessed by solving the missing values in the dataset to enhance the risk prediction process. Experimental results of the proposed method achieved an accuracy of 0.938 for MIMIC-IV dataset when compared to existing methods such as XGBoost and RDF.
Keywords:	Bi-Directional Long Short Term Memory, Disease Risk Prediction, Electronic Health Record, Flying Squirrel Search Optimization, Machine Learning.
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	HYBRID TYPE ERROR CORRECTION CODE TO INCREASE RELIABILITY IN SRAM MEMORIES
Author:	P V GOPIKUMAR , R MANIKANDAN , C RAVI SHANKAR REDDY
Abstract:	This paper presents implementation of mixed mode Error Correction Coder architecture that is capable of detecting and correcting both random and burst. The proposed architecture executes two different types of architectures in which one of them aims at targeting the random while the other aims at burst. The diagnosis of random faults is obtained by employing the modified decimal matrix coding whereas the burst diagnosis is done by flexible unequal error method. The proposed architecture gains an adequate advantage with respect to area overhead and power consumption. The proposed architecture is implemented in cadence. In the front end design the parameters are estimated by using out by GENUS and on similar lines, the back end design is carried out by using out by INNOVUS and the performance metric generated includes Area Report, Power Report, and timing report and chip layout. The area, power and delay overhead of the proposed architecture is found to be 50088.491, 20.7mW and 2.11ps which are very small compared to other ECC that are taken for experimental purpose and this low values of area, power and delay are mainly due to the use of encoder reuse technique. Further the experimental results conclude fact that the proposed architecture can achieve 100% of error correction up to six bits and offers correction rate of 80.76 and 68.43 for burst of length 7 and 8 respectively. The error correction capabilities for random error is found to be 100% for Quintuple Bit Error and there after we observed slight decrement in error correction capability for Sextuple Bit, Septuple bit and Octuple Bit. The percentage of error correction rate for Sextuple bit, Septuple bit and Octuple bit is found to be 46, 4 and 2 respectively. The increased reliability of this hybrid type error correction code is achieved at a cost of slight increment in the overhead bits.
Keywords:	Random errors, Burst errors, Modified Decimal Matrix code (MDMC), Flexible Unequal Error Control (FUEC), and Error Correction Code (ECC).
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	AN EFFECTIVE FILTERING EXTENSION TECHNIQUE OF SMOTE FOR CONTROLLED SYNTHETIC DATA GENERATION
Author:	SOMIYA ABOKADR , AZREEN AZMAN , HAZLINA HAMDAN , NURUL AMELINA NASHARUDDIN
Abstract:	Imbalanced dataset poses a significant classification challenge within the realm of machine learning and has gained growing prominence due to the need to handle real-world data that is usually imbalanced and skewed. Numerous resample techniques have been proposed in the literature to improve the performance by solving the imbalanced problem. But this situation becomes more complex when one class contains a significantly larger number of examples compared to the other classes. Under such circumstances, machine learning-based algorithms can accurately identify examples from the majority class but often struggle or likely fail to recognize instances from the minority class. However, these minority class examples often contain crucial and valuable information. In addition, generating new instances to balance the minority classes by curating some rules by the domain expert as in the standard SMOTE, may not be suitable for some instances leading to misclassification by the model. Therefore, this paper proposed a novel technique namely filter extension of the SMOTE algorithm based on optimised kernel trick on SVM to control the generation of balanced synthetic data in overlapped samples. The main objective of this paper is to predict the minority class instances accurately and robustly addressing the imbalance and overlapping issues. The proposed technique is validated by a rigorous testing framework utilizing a 10-fold cross-validation method to ensure a comprehensive evaluation of support vector machine (SVM) classifier. Several parameters, such as, AUC and G-mean metrics were used to ensure the accuracy, robustness and effectiveness of the proposed technique comparing to other traditional and machine learing-based methods. We experimented with highly imbalanced dataset from KEEL repository. The proposed approach outperformed the standard SMOTE and RC-SMOTE, proven the effectiveness of the proposed approach of filtering imbalance data in improving the generalizability and performance of machine learning classifiers.
Keywords:	SMOTE, Imbalance, Filtering, G-mean, Machine Learning
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	OPTIMIZED GRID LINKED PV SYSTEM WITH ZETA CONVERTER FOR ENHANCED EFFICIENCY
Author:	SRINIVASA RAO JALLURI , GORU RADHIKA , SHASHIDHAR KASTHALA , G RAMAKRISHNA4, NITHYA CHANDRAN
Abstract:	The need for efficient converters is essential for the incorporation of solar PV systems with grid infrastruc-ture to control voltage and maximize energy transmission. Using a Zeta converter and the Incremental Conductance (INC) MPPT algorithm, this study explores a grid-linked photovoltaic scheme. The Zeta con-verter plays a crucial role in stabilizing DC link voltage despite fluctuating solar irradiance, considering a consistent output between 290V and 600V.Through dynamic tracking and maintenance of the PV array's MPP, the INC MPPT approach improves system efficiency. In addition to optimizing energy extraction, this integrated strategy guarantees steady power transmission to the grid, reducing fluctuations and en-hancing overall performance. The proposed scheme demonstrates significant improvements in voltage regulation, energy efficiency, and reliability, making it a viable solution for modern renewable energy ap-plications.However, designing an optimal control strategy for the zeta converter is essential to balance the efficiency, transient response and voltage stability. Its ability to regulate voltage effectively enhances the performance of photovoltaic systems, preventing fluctuations from affecting downstream components. Proposed Simulation outcomes provide the efficacy of the proposed system under dynamic operating conditions, highlighting its effective power transfer. This study provides a comprehensive framework for implementing Zeta converters in Grid-linked PV schemes, bridging the gap between renewable energy gen-eration and grid stability.
Keywords:	Grid-Linked PV Scheme, Zeta Converter, DC Link Voltage, Voltage Regulation, Renewable Energy
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	NEXT GEN BUSINESS INTELLIGENCE: LEVARAGING PREDICTIVE ANALYTICS, AI & REAL-TIME DECISION- MAKING
Author:	AJAI GOPAL BHARTARIYA , S. K. SINGH , AJAY KUMAR BHARTI
Abstract:	In today’s dynamic business landscape, the ability to harness data effectively has become a cornerstone of competitive advantage. Business Intelligence (BI) is evolving rapidly, integrating cutting-edge technologies like predictive analytics, artificial intelligence (AI), and real-time data processing to deliver actionable insights. Predictive modeling plays a pivotal role in modern data-driven decision-making, offering valuable insights into future trends and behaviors across various data sets. This paper investigates the performance of prevalent machine learning models for predictive analytics, with a particular focus on XGBoost. We demonstrate how hyperparameter tuning can significantly enhance the accuracy prediction of XGBoost comparing its performance against other popular models such as decision trees, random forests, and logistic regression. Through a systematic evaluation, we highlighted the effectiveness of XGBoost in capturing complex patterns within the data, resulting in superior predictive performance. Our findings provide evidence that XGBoost, when appropriately tuned, outperforms traditional models, offering a more robust approach for predictive analytics. This research contributes to the ongoing discourse on the application of machine learning techniques, providing practical insights for researchers and practitioners aiming to improve predictive modeling accuracy.
Keywords:	Artificial Intelligence (AI), XGBoost, Random Forest, GridSearch, Business Intelligence (BI), Machine Learning (ML), Data, NLP.
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	ENHANCED PREDICTIVE MODELING FOR ALZHEIMER’S DISEASE: INTEGRATING CLUSTER-BASED BOOSTING AND GRADIENT TECHNIQUES WITH OPTIMIZED FEATURE SELECTION
Author:	S PHANI PRAVEEN , SREEDHAR BHUKYA , SHARMILA VALLEM , SATEESH GORIKAPUDI , KIRAN KUMAR REDDY PENUBAKA , VAHIDUDDIN SHARIFF
Abstract:	The early signs of Alzheimer’s disease (AD) prove difficult to detect since this progressive illness has an ongoing character. The correct prediction of how the disease progresses represents a critical need for providing prompt treatment that enables effective management of the disease. The research develops Cluster-Integrated Boosting and Gradient (CIBG) as a stacked ensemble which combines CatBoost with Gradient Boosting Machine (GBM) classifiers to boost AD diagnostic precision. Kaggle provided the dataset containing diverse health-related characteristics together with lifestyle characteristics and cognitive data. The data preprocessing steps included handling missing values and normalization procedures coupled with SMOTE implementation to solve class imbalance problems. The feature selection process involved using Recursive Feature Elimination (RFE) which determined the most important predictive variables. The CIBG model uses CatBoost’s ordered boosting method to manage categorical features together with GBM’s gradient boosting structure that performs classification. Within CIBG the base classifier chooses logistic regression. Model performance evaluation included accuracy together with precision, recall, and F1-score metrics which reached an optimal rate of 96.8%. The CIBG approach produces results that show better performance than standard classifiers while maintaining stronger robustness and improved interpretability capabilities. The research shows how hybrid machine learning methods help enhance early diagnosis of AD and develops medical systems that are dependable for clinical use.
Keywords:	Recursive Feature Elimination, CIBG, Ensemble Learning Models, SMOTE, Feature Selection, CatBoost, GBM.
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	DATA SEGMENTATION USING MIXTURE REGRESSION MODELS WITH GENERALIZED GAUSSIAN DISTRIBUTION AND K-MEANS
Author:	S A V S SAMBHA MURTHY S , K. SRINIVASA RAO , KUNJAM NAGESWAR RAO
Abstract:	Data segmentation using mixture regression models gained lot of momentum due to their ready applicability in market analytics, business analytics, financial analytics, supply chain analytics , Human Resource analytics etc. In regression analysis it is customary to assume that error term follows a Gaussian distribution. Gaussian distribution has several drawbacks such as being mesokurtic and the model may not serve well for all types of data. Hence, in this paper we develop data segmentation method using mixture of regression models with Generalized Gaussian Distributed (GGD) errors. The GGD includes leptokurtic, platykurtic and Gaussian distribution as particular cases. The model parameters are estimated using Expectation Maximization (EM) algorithm. The initialization of the parameters is done by using k-means algorithm. The data segmentation algorithm is derived using component maximum likelihood under Bayesian framework. The utility of the proposed algorithm is demonstrated with market segmentation. The performance of the algorithm is evaluated by computing segmentation performance metrics such as accuracy, misclassification rate, precision. It is observed that this method performs much better than the earlier data segmentation methods having Gaussian distributed errors for the data sets having leptokurtic and platykurtic response variables.
Keywords:	Segmentation Methods, Regression Analysis, Generalized Gaussian distribution, Market Analytics, Expectation and Maximization Algorithm.
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	ENHANCING CRIME DATA ANALYSIS THROUGH WORD-VECTOR CONVERSION: A RESNIK-GLOVE APPROACH
Author:	SAJNA MOL H S , GLADSTON RAJ S
Abstract:	Crime data analysis is being performed now according to the requirement of Natural Language Processing (NLP) and machine learning algorithms that are able to extract insights from vast amounts of unstructured text data. There have been several efforts to apply classical word embedding models such as Glove, BERT, Skip-Gram, and Bag of Words, but they never identify the explicit meaning of crime tapes. This paper presents a novel hybrid method in which Resnik's similarity measure is combined with the GloVe algorithm to improve the semantic representation of text crime data. Our method employs the Resnik-GloVe model to generate word vectors for crime descriptions and addresses that both capture global word co-occurrence as well as semantic similarity. An Entropy swish based convolutional dense neural network (ES-CDNN) is learned to enhance classifier accuracy by incorporating them. The experiments performed with the 2016 San Francisco crime dataset validate that the Resnik-GloVe method actually generates an outstanding result by obtaining 98.33% accuracy whereas GloVe achieved 96.25%, BERT 94.26%, Skip-Gram 92.15%, and Bag of Words reached as low as 90.66%. The suggested methodology utilizes improved classification of crime data and helps law enforcement officers and policymakers analyze crime patterns more effectively. The research contributes to the area of crime analytics by addressing important shortcomings in existing word embedding methods and demonstrating the real benefits of integrating semantic similarity measures with conventional NLP models.
Keywords:	Crime Data Analysis, Word-To-Vector Conversion, Resnik Glove, Textual Information Representation, Word Embedding Algorithms.
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	ENHANCED SKIN LESION CLASSIFICATION USING DEEP LEARNING MODEL IN INTERNET OF MEDICAL THINGS
Author:	SHANKARA CHIKKALINGAIAH, MAHADEVI KONANAHALLI CHUNCHAIAH , ASHWINI KERAGODU SHIVALINGASWAMY
Abstract:	The recent development of the Internet of Medical Things (IoMT) has greatly benefited medical professionals, patients and physicians in accessing medical information. This information helps to detect and recognize the diseases in the patients through medical images. However, the automatic detection and classification of diseases in medical images like skin lesions, brain tumor, breast cancer, etc., are still challenging in IoMT due to image quality and irrelevant features. To solve this issue, a Logistic Chaotic Map based Red Panda Optimization algorithm (LCM-RPO) is proposed for the feature selection process to select the significant features. Additionally, a Long Short–Term Memory (LSTM) model is utilised to classify the medical images effectively by learning the sequential information. Initially, the medical images are acquired and then raw images are preprocessed to enhance the data quality for further processing. The DenseNet-169 based feature extraction model extracts the most important features from the diseased portion. The experimental results of the proposed LCM-RPO-LSTM model attained an accuracy of 95.40% and 0.973 for Ph2 and ISIC-2016 datasets, which is higher than existing classification approaches such as PLDG and SVM-IARO.
Keywords:	DenseNet-169, Internet Of Medical Things, Logistic Chaotic Mapping Based Red Panda Optimization Algorithm, Long-Short –Term Memory, Medical Image Classification
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	GAN-OST: GENERATIVE ADVERSARIAL NETWORKS FOR PRECISION SEGMENTATION AND SYNTHETIC AUGMENTATION OF OSTEOSARCOMA TUMOURS IN MEDICAL IMAGING
Author:	GONDI HEMAMRUTHA , NAGA MALLESWARI DUBBA
Abstract:	Osteosarcoma is an extremely malignant bone cancer that the accurate segmentation of tumour is important for effective diagnosis and treatment planning. One of the primary challenges facing segmentation models is the paucity of human annotated medical imaging data. In this paper, we introduce a generative adversarial network (GAN) -based method for accurate segmentation and synthesis augmentation of osteosarcoma tumours in medical imaging — GAN-OST. The suggested approach utilizes a U-Net styled generator to achieve precise tumour segmentation and PatchGAN-based discriminator for creation of high-quality synthetic images. To make up for the lack of training data, GAN-OST adopts a conditional GAN to create realistic synthetic tumour images that are difficult to distinguish from real tumours and thus can be used as a supplement for model training so as to improve the generalization of the model. We will evaluate the performance of our model with Dice Coefficient, Intersection over Union (IoU), Sensitivity, Structural Similarity Index (SSIM) and Fréchet Inception Distance (FID). We evaluate our model on these metrics to achieve a comprehensive evaluation of segmentation accuracy and image quality. Experimental results on publicly available osteosarcoma datasets demonstrate the superior performance of GAN-OST relative to traditional segmentation approaches, including a remarkable increase in both segmentation precision and generalization. Additionally, the synthetic data created by GAN-OST successfully compensates for the lack of data in small sets and allows reliable acting together data laden proposal. This work demonstrates the advancement of osteosarcoma tumour segmentation and provides a network used for data augmentation that could greatly help other rare cancer types and multimodal imaging scenarios in future research.
Keywords:	Data Augmentation, Generative Adversarial Networks (GANs), Medical Imaging, Osteosarcoma, Segmentation, Synthetic Data.
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	INVESTIGATION AND COMPARISON OF FEATURE SELECTION TECHNIQUES WITH HYPER PARAMETER TUNING FOR PREDICTING PROACTIVE CARDIOVASCULAR DISEASE
Author:	JANGAM RAGHUNATH , S KIRAN
Abstract:	Despite efforts to treat cardiovascular disease (CVD), yet it remains, still, one of the major cause of death; hence, there has been a need for developing accurate and swift predictive models to enable early diagnosis and intervention for such patients. This study analyzes and considers several feature selection models in the presence of hyperparameter tuning to improve CVD prediction models. Feature selection is important in improving model’s interpretability, reduce computational complexity and remove redundant or irrelevant feature. Additionally, which of the filter, wrapper, and embedded methods (e.g. Mutual Information, Recursive Feature Elimination, and LASSO regression) are best for working with CVD dataset are considered. In order to further improve the model performance, Hyper Parameters tuning on machine learning classifier like Logistic Regression, Support Vector Machine, Random Forest and XGBoost using Grid Search and Bayesian optimization are applied. Different feature selection and hyper parameter tuning combination are assessed based on its performance metric including accuracy, recall, precision, F1-score and area under receiver operating characteristic (ROC-AUC) curve. The proposed method has attained 89% which is far better than the models like Logistic Regression (LR) of 85%, Support Vector Machine (SVM) of 81%, K-Nearest Neighbors (KNN) of 86.9%, Artificial neural networks (ANN) of 87%. In particular the proposed method is 4.7% better than LR, 9.8% better than SVM, 2.29% better than KNN and 2.3% better than ANN. The results from experiment show that the performance of model can be improved greatly with the complement of XGBoost and Random Forest when feature selection integrated with optimized hyperparameter tuning. Modern IT methodologies boost efficiency of models by using fewer resources while achieving extended operational life. Feature selection integration enhances diagnostics predictions because IT-based analytics plays an important role in medical diagnosis.
Keywords:	Cardiovascular Disease, Feature Selection, Hyper-parameter Tuning, Machine Learning, Predictive Modeling.
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	THE MAPPING AND IDENTIFICATION MODEL OF IT TECHNOPRENEURSHIPS STUDENT POTENTIAL USING DATA MINING AND ARTIFICIAL INTELLIGENCE
Author:	WIJI LESTARI, SINGGIH PURNOMO, INDRA HASTUTI, SRI SUMARLINDA
Abstract:	The Information Technology (IT) Technopreneurships are growing very rapidly in line with technological increasing. This study aims to produce a model for mapping and identifying of student potential in IT (information technology) technopreneurships using intelligent computing. This mapping is based on the existence of 4 areas of IT technopreneurship, namely Software Application developer, Data Analyst, Computer System & Network engineer and Multinedia& Graphics Developer. The indicators used are: interesting, entrepreneurial values, learning styles (Solitary, Logical, Aural, Social, Verbal, visual and Physical) and multiple intelligences (Visual Spatial, Logical Mathematics, Body Kinesthetics, Naturalist, Musical, Linguistics, Interpersonal and Intrapersonal. This study used research and development and action research approach methods. The development of an intelligent computing system used the prototype method. The resulted of internal testing of running well models with cyclometic complexity was lower. The novel model that has been produced is a mapping based on students personal cognitive characteristics for IT technopreneurship based on intelligent computing. User acceptance Test resulted 96%, Expertise judgment Test (Computing / IT) yields 98% and Expertise Judgment Test (Entrepreneurships / Technopreneurships) yields 95%. The result of model validation of 20 IT technopreneurships was precise 85%. The resulted of the implementation of the model resulted in 60 students of Software Application developer, 65 students of Data Analysis, 56 students of Computer Systems & Network engineers, 50 students of Multinedia& Graphics Developer and 24 students not suitable to become IT technopreneurs.
Keywords:	Mapping And Identification Model, Students Potential, IT Technopreneurships, Intelligent Computing, Personal Characteristics
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	PROTOTYPING OF A TWO-WHEELED MOBILE ROBOT FOR SUSTAINABLE MANUFACTURING DEVELOPMENT BASED ON TRIANGULATION METHOD AND SOFTWARE DEVELOPMENT
Author:	MOHAMMAD HAMDAN , ISRAA WAHBI KAMAL , AMER ABU-JASSAR , SVITLANA MAKSYMOVA , VYACHESLAV LYASHENKO
Abstract:	This article presents a mobile robot prototyping and software development for its control production. A created robot prototype moves using a two-wheeled base and is equipped with ultrasonic sensors. This design allows increasing the maneuverability of the robot and reducing the turning radius. The specific arrangement of the sensors allows expanding the area of obstacle detection. The developed software includes receiving data from the sensors, processing them and constructing a movement trajectory in accordance with the target and the current environment state. The trajectory is constructed using the triangulation method. This allows accurately determining the distance to the obstacle, constructing a rational movement trajectory and going around obstacles that may arise on the path of the robot. A number of experiments were conducted that allow us to say that the obstacle detection range is sufficient for a timely response and change the trajectory of movement.
Keywords:	Mobile Robot, Triangulation, Ultrasonic Sensor, Two-Wheeled Robot, Software for Mobile Robot
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	EVALUATION OF ADHD: CLASSIFICATION, TREATMENT STRATEGIES IN PEDIATRIC AND ADOLESCENT POPULATIONS USING MACHINE LEARNING
Author:	M GEETHA PRIYA , P KALYANARAMAN
Abstract:	Attention-Deficit/Hyperactivity Disorder (ADHD) is a neurodevelopmental disorder characterized by inattention, impulsivity, and hyperactivity. Despite extensive research, existing diagnostic approaches rely heavily on subjective clinical assessments, leading to misdiagnoses and delayed interventions. Furthermore, conventional treatment strategies lack personalization, often resulting in suboptimal therapeutic outcomes. This study addresses these gaps by leveraging machine learning techniques to enhance ADHD classification and treatment strategies, specifically for pediatric and adolescent populations. Unlike previous studies that primarily focus on symptom-based categorization, our research integrates multi-modal data, including neuroimaging, behavioral assessments, and genetic markers, to improve diagnostic accuracy. A novel hybrid machine learning model is proposed, incorporating convolutional autoencoders and advanced neural architectures to extract discriminative features for precise classification. Additionally, the study explores AI-driven personalized treatment recommendations, optimizing intervention strategies based on patient-specific patterns. The findings of this research contribute to the development of an objective, data-driven framework for ADHD diagnosis and treatment, reducing reliance on subjective evaluation. This work establishes a foundation for future AI-assisted clinical decision-making, ultimately improving patient outcomes and advancing ADHD research.
Keywords:	ADHD, Brainwave, CNN, Cognitive, Disorder
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	ENHANCING STOCK MARKET INVESTMENT DECISIONS THROUGH BLOCKCHAIN TRANSACTION SECURITY: A STUDY ON INVESTOR INTENTIONS
Author:	RAYMOND HARYADI, ELFINDAH PRINCES
Abstract:	This study analyzes the effect of transaction security using Blockchain technology on customers' decision to invest in the stock market. Blockchain is known as a revolutionary technology that offers high security through decentralization, transparency, and immutability. In the context of investment, this technology has the potential to reduce the risk of data manipulation, fraud, and information leakage that are often a challenge in online investment.A quantitative approach is used with the Theory of Planned Behavior (TPB) framework to measure the influence of variables such as attitude towards money, subjective norms, perceived behavioral control, and Blockchain transaction security on investment intentions. The sample consisted of 460 respondents who had experience or interest in investing in listed companies. Data was collected through a structured questionnaire and analyzed using Smart PLS to validate the research model.The results show that Blockchain-based transaction security, such as cryptographic encryption, smart contracts, and transparent transaction recording, has a significant influence on investment intention by increasing investor confidence. Psychological factors, such as attitudes toward money and subjective norms, also strengthen investment intentions, reflecting the importance of perceived security and social support. This research makes a practical contribution for companies to adopt Blockchain as a strategic move to increase investor confidence. This technology can be the basis for creating a safe and sustainable investment ecosystem. In addition, this research enriches the academic literature on the role of Blockchain in investment decisions and recommends better investment security standards for regulators.
Keywords:	Blockchain, Transaction Security, Stockmarket, Theory Of Planned Behavior (TPB), Intention To Invest, Transparency, Risk Of Fraud, Behavior Control, Subjective Norms.
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	MULTI-LEVEL IMAGE DENOISING INTEGRATED MORPHOLOGY-BASED IMAGE QUALITY ENHANCEMENT MODEL WITH EDGE-BASED SEGMENTATION FOR ACCURATE PANCREATIC CANCER DETECTION
Author:	SRIPATHI CHAITANYA BHARATHI , ESWARAIAH RAYACHOTI
Abstract:	Despite the availability of various methods, diagnosing and treating Pancreatic Cancer (PC) continues as a significant challenge across all types of tumours with its asymptomatic development. Among the most devastating diseases, pancreatic cancer has claimed the lives of countless people around the globe. Traditional methods of diagnosis relied on manual analysis of massive datasets, which was laborious, error-prone, and time-consuming. Therefore, CADs, which use machine learning and deep learning techniques for pancreatic cancer denoising, segmentation, and classification, become necessary. However, there are obstacles to medical image analysis of pancreatic cancer owing to vague symptoms, high rates of misdiagnosis, and substantial monetary expenses. A potential answer is Artificial Intelligence (AI), which can reduce patient expenditures, improve clinical decision-making, and alleviate the workload of medical workers. The use of medical imaging scans has allowed many cancer patients to detect anomalies earlier on. This research makes use of Computed Tomography (CT) images for performing image denoising and segmentation. The high price of the required equipment and infrastructure makes it difficult to spread the technology, which means that many people cannot afford it. Pancreatic cancer detection using medical image analysis is greatly impeded by noisy, low-quality images that mask important diagnostic details and lower detection accuracy. This research tackles the important problem of creating a state-of-the-art image processing method that can reliably and effectively denoise medical images, improve image quality using morphological techniques, and apply precise edge-based segmentation to better visualize and detect pancreatic cancer early on. In this research, a new Multi-Level Image Denoising and Integrated Morphology based Image Quality Enhancement Model (MIDIMIQEM) by edge-based segmentation for precise pancreatic cancer detection is proposed. This proposed model solve these problems in medical image analysis by introducing multi level denoising technique and also morphology-based enhancement with a robust edge based segmentation for an enhanced diagnostic accuracy. The model uses wavelet based multilevel image denoising for the removal of noise and thereafter morphological operations are employed to improve the contrast which helps in easier differentiable identification of tumor structures. The segmented regions are further improved by the edge detection techniques, which increases the chances of self and accurate determination of cancerous tissues. Experimental evaluation on MIDIMIQEM demonstrates that the image quality, segmentation accuracy and small nodules detection performance are significantly improved compared with state-of-the-art models. The proposed model achieved 98.7% accuracy in Multi-Level Image Denoising and 99.3% accuracy in Segmentation Accuracy. This novel technique is expected to help radiologists make an early and accurate diagnosis that could improve patient outcomes. With the use of cutting-edge edge-based morphological processing methods, the suggested multi-level picture denoising and morphology-based enhancement model outperforms state-of-the-art methods for pancreatic cancer image segmentation, greatly enhancing diagnostic precision and opening up new avenues for early detection.
Keywords:	Image Denoising, Morphology-Based Enhancement, Pancreatic Cancer Detection, Edge-Based Segmentation, Medical Image Analysis, Wavelet Transform, Tumor Localization.
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	ENHANCING INTER-DOMAIN ROUTING POLICY AUTOMATION WITH BLOCKCHAIN AND SMART CONTRACTS
Author:	SANKARA MAHALINGAM M, N. SURESH KUMAR , R. KANNIGA DEVI
Abstract:	The Border Gateway Protocol (BGP), which manages inter-domain routing, undergoes frequent changes. This requires strong policy management to keep data transfer across different networks (called Autonomous Systems, or ASes) safe and efficient. Traditional methods usually require physical setup, which can cause mistakes and security risks. This paper looks at how blockchain technology and smart contracts can work together to make inter-domain routing rules automatic. We propose a framework called the RPAChain (Routing Policy Automation Chain), which enhances trust, transparency, and automation in routing policy management by utilising the decentralised and immutable ledgers of blockchain technology along with the self-executing capabilities of smart contracts. We explain the system's working details, including the automatic routing processes. We then examine the results and their implications for future routing between various domains.
Keywords:	Inter-Domain Routing, Border Gateway Protocol (BGP), Blockchain, Smart Contracts, Autonomous Systems (ASes), Routing Policy Automation, Network Security.
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	AI-POWERED EMPATHY: SENTIMENT ANALYSIS IN PERSONAL CARE USING RoBERTa AND XLNet
Author:	DR. REVATHI DURGAM , NARENDRA BABU PAMULA , NALLANI DHARANI ,DR. B V N SIVA KUMAR ,DR. PUTTA DURGA , VELCHURI BALAJI
Abstract:	Since they reveal vital information about consumer preferences and satisfaction, personal care products in the current digital era depend especially on user reviews. The intricacy and diversity of natural language make large-scale evaluation of these tests challenging. This study is motivated by the need of the personal care industry to discover how sentiment analysis grounded on Natural Language Processing (NLP) techniques could be useful. We want to use NLP models to quickly and more precisely detect consumer sentiment, so guiding classification. From this, companies can choose a lot of knowledge on consumer attitude and product performance. Covered include how to select appropriate models, compile sentiment analysis review data, and how sentiment classification influences marketing and product strategy. In the increasingly competitive personal care industry, natural language processing (NLP) can help to streamline user input thereby facilitating data-driven decisions and greater consumer satisfaction. People are realizing the full possibilities of Advanced Technology, which marks a period of perpetual and never-ending advancement. In the field of shopping and virtually purchasing goods using technology dubious issues of product quality surface. Reviews give us direct answers and allow us to sort through our questions. People are asked to post reviews, in which they provide honest and forthright comments, following purchases of goods. Our objective is to use Sentimental Analysis with NLP to classify the past reviews of the product as either favourable or negative with accuracy of 94%. Customers as well as the product company will benefit from this in order to make appropriate judgments and apply correct modifications.
Keywords:	Natural Language Processing (NLP), Sentiment Analysis, Customer Reviews, Personal Care Products, Opinion Mining, Text Classification, Emotion Detection, Product Feedback Analysis, Consumer Insights
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	DEVELOPMENT OF NEW MACHINE LEARNING ALGORITHM FOR CUCUMBER AND GRAPE LEAF DISEASE DETECTION USING MULTI SVM WITH CUSTOM KERNELS
Author:	C.NANCY , DR.S.KIRAN
Abstract:	In Agricultural production crop disease is a wide spread problem, it effects the productivity and quality of crops. Cucumbers are high water content vegetables and many minerals and vitamins, But Cucumbers are susceptible to several diseases. In wine production grapes are main ingredient, but these are severely attack by brown spot, mites, Anthracnose, downy mildew, black rot, and leaf blight. Identification and detection of crop diseases are at most care for reducing economic loss for the formers also improve productivity. But manual detection takes huge amount of time, sometimes it gives inaccurate results. To overcome these challenges by using advanced artificial intelligence techniques including Alex net, VGG-16,U-Net,ResNet,VGG-19,Fine KNN, Random forest Algorithms, YOLO v5.But these algorithms struggle with irrelevant features, noise and poor performance. To handle these we proposed novel approach using Multi SVM with Custom SVM kernels along with the GLCM features. The primary objective of the new algorithm is to enhance the model’s performance and increase the efficiency of disease detection. Our datasets cucumber and grapes collected from Kaggle and directly from cucumber farms, it contains four different categories: Healthy, Powdery mildew, Downy mildew and Target Leaf Spot. First phase is to apply image augmentation technique for enhancing the image. Second affected area is calculated using K-Means Algorithm. Next from the images GLCM features are extracted. Subsequently classification is performed using Multi SVM algorithm with Custom SVM kernels. Our Novel Approach achieved best identification result with mAP is 90.62% compare with YOLO v5M mAP is 84.6.
Keywords:	Random Forest Algorithms, YOLO v5, GLCM, mAP, MultiSVM.
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	SURVEY ON SOFTWARE ARCHITECTURE FOR QUANTUM COMPUTING: DESIGN PRINCIPLES, CHALLENGES, AND FUTURE DIRECTIONS
Author:	KHALIL JAMOUS
Abstract:	Quantum computing is imagined as the route to moving the current paradigm for problem solving that could not, under the old paradigm, be solved within a feasible period of time. In spite of these promises, it needs to fulfill all of these by demonstrating the proper architecture upon which such processing challenges that the quantum computer hardware dictates could be mounted. This paper revisits these very basics forming the pivot of design quantum software architecture and is guided by principles relating to modularity, hybridization, scalability, and fault tolerance. A focus on these, but putting much emphasis on modularity and hybridization on issues relating to handling the complexity or optimizing the system's performance. The paper also highlights some major challenges in the development of quantum software: hardware constraining, lack of standardization, and the need for effective fault tolerance mechanisms. The survey will provide comparisons between existing frameworks like Qiskit, Cirq, and Amazon Braket to carve out interoperability gaps and platform specialization. It also discusses some of the future directions, including hybrid architectures, integration of machine learning, standardization efforts, and co-design of hardware-software systems that will be essential to drive quantum computing toward scalability and widespread adoption. The work is a rich source for both the researcher and the practitioner by setting out an overview of the current status and offering a vision as to how to create the next generation quantum software architectures, which will be efficient, scalable, and prepared to respond to diverse application demands.
Keywords:	Quantum Computing, Software Architecture, Design Principles, Challenges, Future Directions.
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	A NEW PARADIGM IN FRAUD DETECTION: LEVERAGING SOCIAL MEDIA TO PREDICT MCCS IN REAL TIME
Author:	MAROUANE AIT SAID , ABDELMAJID HAJAMI , AYOUB KRARI
Abstract:	Card-not-present (CNP) fraud presents a persistent challenge in the financial industry, costing institutions billions annually. Traditional fraud detection models rely heavily on transaction history and rule-based mechanisms, making them reactive rather than proactive. These systems struggle to detect emerging fraud patterns that evolve beyond historical transaction-based limitations. In this study, we introduce an innovative fraud detection framework that leverages social media analytics to predict a cardholder’s likely future Merchant Category Code (MCC), providing a proactive layer of fraud prevention. Our approach integrates natural language processing (NLP) techniques with machine learning to analyze publicly available social media posts from business pages linked to specific MCC categories. By transforming unstructured text data using Term Frequency-Inverse Document Frequency (TF-IDF) vectorization, we extract relevant linguistic features and apply a Multinomial Naïve Bayes classifier to predict MCCs with high accuracy. To ensure model scalability and robustness, we expanded our dataset from 120,000 to over three million records using data amplification techniques. Our model achieves an accuracy of 99.1% through cross-validation, significantly improving real-time MCC prediction for transaction authentication. Unlike conventional fraud detection systems that react only to anomalies in past spending behavior, our method proactively forecasts spending categories based on dynamic social media interactions. By comparing the predicted MCCs with those in real-time transactions, financial institutions can identify potential fraudulent activities before they occur. This novel integration of behavioral insights from social media into fraud detection enhances the adaptability of fraud prevention systems, bridging the gap between static transaction monitoring and real-time fraud anticipation. This research demonstrates that incorporating social media discourse into financial security mechanisms can revolutionize CNP fraud detection. Future work will focus on expanding the model to accommodate multilingual data and additional MCC categories, ensuring its applicability across diverse consumer markets.
Keywords:	Card-Not-Present Fraud, Machine Learning, Merchant Category Codes (MCC), Multinomial Naive Bayes Classifiers, Natural Language Processing (NLP), Predictive Analytics, Proactive Fraud Prevention, Real-Time Fraud Detection, Merchant Category Code (MCC).
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	USING VIDEO TITLES TO PREDICT YOUTUBE AUDIENCE BEHAVIOR BASED ON MACHINE LEARNING
Author:	CHIH-CHIEN WANG , YI-FENG LIN , YA-CHEN HSIEH , YU-HAN KAO
Abstract:	As the leading video-sharing platform, YouTube has evolved into a new form of media, facilitating content dissemination by creators and fostering active audience participation. Enhanced viewer engagement offers creators increased influence and financial benefits due to heightened popularity. However, the vast quantity of videos available on YouTube means that only a select few manage to captivate significant audience attention. Within YouTube's interface, video titles play a crucial role in attracting users, making the exploration of the relationship between audience engagement and video titles an important research topic. This study aims to predict audience engagement by analyzing video titles using various machine learning methods. The study employs the Linguistic Inquiry and Word Count (LIWC) software for natural language processing, which calculates the frequency of word categories in the text, such as pronouns, emotional, and cognitive words, presenting these frequencies as relative percentages. Results indicate that different textual categories within video titles correlate significantly with audience liking and commenting behaviors. Among the evaluated machine learning techniques, Random Forest and K-Nearest Neighbors Regression models exhibit superior predictive performance. The findings provide insights into the intricate interplay between textual features of video titles and their impact on audience engagement. This research serves as a valuable reference for video creators, guiding their decision-making process concerning video titles to optimize audience interaction. Additionally, the study contributes to existing literature by elucidating the nuanced relationship between textual elements in video titles and audience responsiveness.
Keywords:	Video titles, Linguistic Inquiry and Word Count (LIWC), Machine learning, Audience behaviors
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	A NOVEL ENTROPY-BASED CASCADED CAPSULE NEURAL NETWORK WITH AN OPTIMIZED LSTM FOR ANOMALY SEGMENTATION AND CLASSIFICATION
Author:	SHAMEEM AKTHAR K, DR. K. LAKSHMI PRIYA
Abstract:	The topic at hand of Anomaly Detection (AD) is important and well-researched. But creating successful AD techniques for complex and high-dimensional data is still difficult. There exist various methods for the detection of anomalies. Though, there are various existing methods, to improve the efficiency and to bring more advantages a method called Entropy based cascaded capsule neural network (CCNN) with an optimized LSTM (OLSTM) for segmentation and classification is suggested for the AD in the videos. Initially, pre-processing is done using the Gaussian filter followed by the segmentation using the Cascaded Capsule Neural network. Using this segmented data along with the dataset classified using entropy, Feature extraction (FE) is done by which the features are extracted. Finally, classification is done using the Optimized LSTM using ROCO algorithm method. The suggested approach has an accuracy of 98%, specificity of 99.8%, sensitivity of 92.9%, Precision of 92.9, F measure of 92.9%, FNR of 10%, FPR of 0.3%and Matthew Correlation Coefficient (MCC) of 89%. Thus, from the results, it is seen that our suggested approach performed more effectively than the other methods that are currently in use.
Keywords:	Anomaly Detection, Entropy, CCNN, OLSTM, ROCO
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text

Title:	ADAPTIVE TRAFFIC SIGNAL CONTROL USING AI AND DISTRIBUTED SYSTEMS FOR SMART URBAN MOBILITY
Author:	ISMAIL ZRIGUI , SAMIRA KHOULJI , MOHAMED LARBI KERKEB
Abstract:	Urban traffic jams remain a pressing issue due to increasing vehicle concentrations, inadaptive traffic signal management, and unreliable road conditions. Fixed or semi-adaptive traditional traffic control systems cannot adaptively change in real-time traffic environments, leading to wasted time, excessive fuel usage, and greater pollutant emissions. This study proposes a comprehensive framework for the optimization of urban traffic, leveraging predictive modeling, adaptive signal control, and distributed messaging infrastructure. The platform collects real-time sensor information, employs machine learning techniques for short-term traffic prediction, and optimizes signal timing through heuristic approaches such as Simulated Annealing (SA) and Reinforcement Learning (RL). The asynchronous messaging architecture supports flexible communication among prediction modules, controllers, and external services to facilitate flexibility and scalability. Experimental validation was conducted using the SUMO traffic simulator, where a standard fixed-cycle signal scheme was compared to the proposed adaptive scheme. Results indicate that there is a 21.6% reduction in average waiting time, a 12.9% reduction in CO₂ emissions, and a 15.6% improvement in fuel efficiency. These findings validate the effectiveness of the system in mitigating congestion and enabling green urban mobility. Future work will involve incorporating vehicle-to-everything communication data (V2X communication), multi-intersection coordination, and deep learning traffic flow predictions in order to better enhance adaptability and robustness at the large-scale deployment stage.
Keywords:	Intelligent Transportation Systems, Traffic Optimization, Adaptive Signal Control, Predictive Modeling, Reinforcement Learning, Smart Cities
Source:	Journal of Theoretical and Applied Information Technology 30^th April 2025 -- Vol. 103. No. 8-- 2025
Full Text