		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
August 2025 | Vol. 103 No.16

Title:	FUZZYFAKEROBERTA: FAKE REVIEW IDENTIFICATION IN E-COMMERCE PLATFORM
Author:	TEH NORANIS MOHD ARIS, KHAIRUNNISA ABDUL RAMAN, AZREE SHAHREL AHMAD NAZRI
Abstract:	Fake reviews on e-commerce platforms pose a substantial risk to the integrity of online reviews and can significantly mislead consumers, leading to unfavorable buying choices. These fake reviews can distort consumer perceptions, potentially resulting in financial losses and decreased trust in online shopping. This study investigates the integration of fuzzy logic with the fine-tuned RoBERTa model, resulting in the "fuzzyfakeRoBERTa" model, designed to detect fake reviews on e-commerce platforms. The fuzzyfakeRoBERTa model enhances the accuracy and precision of fake review identification by effectively addressing the inherent imprecision and uncertainty often present in data. The research methodology included replicating the original fakeRoBERTa model, incorporating fuzzy logic to handle ambiguous data better, and thoroughly evaluating the model's performance using key metrics such as accuracy, precision, recall, and F1-score. The fuzzyfakeRoBERTa model achieved a notable accuracy rate of 97.59%, indicating a significant improvement over the original model's performance. This enhanced accuracy demonstrates the model's superior robustness and effectiveness in identifying fake reviews. The findings suggest that integrating fuzzy logic into deep learning models can substantially improve their performance in tasks that involve complex and nuanced data. This research enhances the reliability and credibility of e-commerce platforms by offering a more precise and effective tool for detecting fraudulent reviews, thereby helping maintain consumer trust and ensure fair competition in the digital marketplace.
Keywords:	Fake Reviews, E-Commerce, Fuzzy Logic, Fakeroberta, Machine Learning, Deep Learning, Text Classification
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	TORQUE RIPPLE MINIMIZATION FOR SPM AND IPM DRIVE USING FUZZY AND PI CONTROLLERS WITH AN ONLINE STATOR REFERENCE FLUX
Author:	N. KRISHNA KUMARI, R. GESHMA KUMARI, K. SRAVANI
Abstract:	The torque ripples are crucial to the drive's ability to operate dynamically well. The torque ripples cause speed fluctuations that impair the drive's performance. The performance of PMSM drives is deteriorated, especially at low speeds, by these torque pulsations, which fluctuate periodically with rotor position and manifest as speed ripples. Therefore, a drive's successful dynamic functioning depends on the torque's smooth change. Direct Torque Control (DTC)'s primary characteristic is its lack of accurate mathematical models. Even though DTC's decoupling feature allows it to manage torque and flux independently, it can be challenging to choose the precise voltage space vector, which leads to substantial torque and flux ripples. However, with the aid of an appropriate voltage space vector, it is crucial to maintain torque and flux within certain bounds. Three sets of DTC space vectors are used to investigate the drive's transient and steady state performance using the "Maximum Torque per unit Current/Ampere (MTPA)" operation. Instead of using stator flux linkage, the online stator reference flux is determined using the "A Maximum Torque per Ampere (MTPA)" approach, which uses quadrature stator flux, direct stator flux, and torque reference. The performance of the three control parameters is assessed in terms of flux ripple, torque ripple and transient responsiveness to step variations in torque and speed control instructions. Three sets of DTC space vectors are taken into consideration with MTPA operation for analytical purposes. By selecting the appropriate membership functions and rule base in a fuzzy logic controller (FLC), torque and flux ripples can be significantly reduced, resulting in an accurate voltage space vector with a smaller hysteresis band. An attempt has been made to simulate and analyse DTC on PMSMs with both SPM and IPM rotor configurations. The performance of IPM and SPM are compared using DTC Strategy, by simulation with MATLAB/Simulink Tools. This paper's goal is to use DTC-SVM to reduce torque ripples of a PMSM drive on both IPM and SPM rotor designs. A drive's dynamic performance can be examined offline with the use of simulation software, which lowers costs and time and enables real-time system design.
Keywords:	MTPA, FOC, DTC, PI Controller, Fuzzy Logic Control
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	ARTIFICIAL INTELLIGENCE DRIVEN OPTIMIZATION OF PV STATCOM PI CONTROLLERS FOR REGULATORY COMPLIANT POWER QUALITY IN MOROCCAN DISTRIBUTION NETWORK CONTEXT
Author:	SAAD SARIH1, SAMIRA CHABAA , ZAKARIA BOULGHASOUL , ABDELOUAHED TAJER, ABDELHADI EL BACHA
Abstract:	The present paper leverages artificial intelligence algorithms to optimize and tune the parameters of PI controllers to enhance the performance of a PV STATCOM in terms of both dynamic and static responses. The proposed approach's effectiveness has been validated using a distribution grid model feeding capacitive and inductive loads at the point of common coupling (PCC) and controlled by three PI regulators. Comparative analysis between the traditional PI controllers and the refined parameters through Particle Swarm Optimization (PSO) and genetic algorithm (GA) reveals that this method substantially enhances both the static and dynamic performance characteristics of the whole system and allows providing dynamic grid support in faulty conditions. In addition, it ensures power factor correction while injecting active power. The study is performed and tested under technical Moroccan regulations. The STATCOM is implemented between a PV array module and the PCC to the distribution network, which is typical of a standard structure in the presence of renewable generators. The PV STATCOM is not yet used in current Moroccan electrical network context and the proposed system effectively manages network variability and grid disturbances, notably enhancing power quality at the (PCC). It dynamically adjusts to supply or absorb reactive power as needed, all within an acceptable reversing time, ensuring stability and resilience in the network. MATLAB Simulink software is used to conduct simulation, which provides the best outcomes while complying to the technical constraints set by local regulations.
Keywords:	Moroccan regulations, PSO, GA, PCC, PV-STATCOM, Power Factor correction, Reactive power compensation.
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	THE IMPACT OF DIGITAL VIDEO SURVEILLANCE AND ANALYSIS TECHNOLOGIES IN THE INVESTIGATION OF CRIMES AGAINST THE PERSON
Author:	DMYTRO BUCHENKO, OLENA KALHANOV, HANNA STEPANOVA, VITALII MATSAK, VASYL SMIKH
Abstract:	The impact of digital video surveillance technologies in the investigation of crimes against the person has become increasingly significant due to the integration of advanced video analysis tools in modern criminal justice. This study evaluates the role of video surveillance, particularly digital video analysis, in enhancing the evidentiary process during investigations of enforced disappearances. The research applies a combination of qualitative and quantitative methods, including legal dogmatics, case law analysis, comparative legal approaches, legal monitoring, and mathematical modelling. Findings reveal that video surveillance technologies contribute crucially to criminal investigations, offering objective visual evidence that complements or challenges testimonial accounts. Notably, statistical analysis demonstrates that video footage supported the resolution of 72% of general crimes against the person, with this figure rising to 86% in enforced disappearance cases. Furthermore, improving video quality by 20% was shown to increase the probability of case resolution by 15%. Despite these benefits, the study highlights ongoing privacy concerns and emphasizes the need for robust legal frameworks to safeguard rights and ensure the admissibility of digital video evidence. Future research should investigate the integration of artificial intelligence (AI) into video analytics to further enhance investigative outcomes.
Keywords:	Video Analysis; AI-Driven Video Analytics; Criminal Investigation; Enforced Disappearances; Evidentiary Value
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	AUTOMATED ASSESSMENT SYSTEMS IN CLOUD ENVIRONMENTS FOR THE FORMATION OF FUTURE PROFESSIONALS’ DIGITAL COMPETENCE
Author:	PAVLO PAVLYHA, VOLODYMYR DEHTIAROV, TETIANA HERASYMCHUK, MARYNA VYSHNEVSKA, KSENIIA KUGAI
Abstract:	The relevance of the study stems from the imperative to enhance the digital competencies of future professionals through the implementation of contemporary assessment technologies. The purpose of this research was to examine the impact of automated assessment systems within a cloud-based environment on the development of students’ digital skills. The investigation used an experimental method, incorporating online assessments utilizing the adapted DigComp 2.1 framework, alongside questionnaires and content analysis. The control group engaged studied using traditional methods, while the experimental group utilized a cloud-based system based in Google Workspace: Google Classroom for course management, Google Forms with automatic assignment checking, and Google Sheets for analysing results. The results showed that digital competence in the experimental group increased by 9.0 points (from 60.2 to 69.2), in the control group — by 3.2 points (from 59.2 to 62.4). The largest increase in the experimental group was recorded in digital communication, content creation, and safety in the digital environment, with each area reflecting an impressive increase of 9 points. According to the survey, 87% of students felt confident in working with cloud services, 83% realized their strengths and weaknesses, and 81% rated quick feedback as a motivating factor. It is concluded that cloud-based automated assessment effectively promotes the development of digital skills: self-control, communication, analytical thinking, and confidence. The scientific novelty lies in the comprehensive analysis of the automated assessment’s impact on various components of digital competence. Prospects for further research include the integration of adaptive systems and the study of the long-term effects of digital platforms on learning outcomes.
Keywords:	Automated Assessment, Cloud Technologies, Digital Competence, Future Professionals, Higher Education, Educational Platforms, Information Technologies.
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	INTEGRATED ENCRYPTION FRAMEWORK FOR ENHANCED DATA SECURITY IN CLOUD STORAGE
Author:	P HIMA BINDU DR. B. HARICHANDANA DR.BHASKARREDDY T
Abstract:	In the contemporary era, there has been a consistent increase in the usage of cloud infrastructure to manage enterprise data. However, there have been security concerns from the data owners’ point of view, no matter how good the security of the Cloud Service Provider (CSP) is claimed. Security mechanisms based on cryptography, such as AES, DES, and RSA, have been serving information systems for safeguarding data from adversaries. Although they are effective at protecting data from malicious attacks, future-proof security mechanisms must also consider post-quantum threats. Many of the current studies predicted the possibility of post-quantum security vulnerabilities of existing cryptographic primitives. Motivated by the findings from the literature on the need for novel mechanisms to deal with security of cloud data in post-quantum era, we proposed a novel security framework known as Cloud Data Security Framework (CDSF) that has underlying mechanisms for encoding and decoding that are used to have stronger security when compared with traditional cryptographic primitives such as AES for cloud data encryption and decryption. Moreover, the framework achieves data security, data availability, and data integrity with its mechanisms used for encoding and decoding data. Data owners can use the proposed framework to have secure outsourcing of data and data retrieval even in the post-quantum era. The proposed framework is evaluated with an empirical study using Amazon EC2 cloud infrastructure and found to have better performance over the state-of-the-art cryptographic primitives like AES, DES, and RSA.
Keywords:	Cloud Computing, Cloud Data Security Framework, Security, Encoding, Decoding
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	ANALYSIS OF E-COMMERCE PRODUCT REVIEW FOR TRUST ASSESSMENT USING ENSEMBLED DEEP LEARNING MODEL
Author:	A.SHALINI, DR.R.ROOPA CHANDRIKA
Abstract:	In the realm of e-commerce, online reviews plays a vital role in influencing consumer decisions.. Especially there is huge demand in computing the trust of the online reviews to prevent the consumer from misguiding and major impacts of the products of the similar businesses. However many researcher has initiated their trust assessment approaches in online product review using machine learning and deep learning based Artificial intelligence architecture. Despite of several benefits, those approaches faces major challenges in terms of the processing the linguistic features and semantic features. In this paper, a new ensemble deep learning model with sentiment analysis has been designed to evaluate trust in product reviews from the Flipkart platform on providing consumer with reliable deeper insights of the product quality and user experiences. Initially preprocessing of the dataset is carried out to obtain reliable word vector from product reviews containing text data through stop word removal , special character removal , stemming and tokenization processes. Next, N-gram model is used to extract the contextual and behaviour features through linguistic and semantic approaches on token of the word vector. Obtained contextual feature and behaviour word vector is encoded using transformer model. Encoded Word vector is projected to BERT analysis for sentiment extraction. Encoded word vector with sentiment is projected to ensemble deep learning models termed as EDL-RNN+CNN+BIGRU. It composed of Recurrent Neural Networks (RNN), Improved Pooling based Convolutional Neural Networks (IPCNN) and Bidirectional Gated Recurrent Units (BiGRU) model. Finally the complementary strengths of each model is leveraged as it is more robust and effective in capturing the nuanced features of online reviews for sentiment analysis compared to using any single architecture alone. Weighted averaging is a common approach employed to ensemble learning to interface the predictions of individual architectures while assigning different weights to each model's prediction based on its performance on computing the trust on the processing sentiment incorporated encoded feature vector in the layer of the model. Experimental and performance analysis of the proposed ensemble model is carried out using Flipkart dataset which is extracted using the online platform such as kaggle. Performance analysis of the model using cross fold validation of test data is performed to compute the accuracy and effectiveness of the model . On evaluation, model produces 94.2% of accuracy, 94% of precision and 93% of recall which found to be outperforming on comparing it against state of art approaches.
Keywords:	Ensemble Deep learning, RNN, CNN and BGRU.
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	AI-DRIVEN FRAMEWORK FOR EARLY DETECTION OF PLANT STRESS USING MULTI-SOURCE REMOTE SENSING DATA AND MACHINE LEARNING TECHNIQUES
Author:	L K SURESH KUMAR, RAVI UYYALA, ADUSUMILLI RAMANA LAKSHMI, BUJJIBABU PENUMUTCHI, SHARMILADEVI B, CH. SABITHA, DIVVELA SRINIVASA RAO
Abstract:	In this article, we suggest an approach of machine Learning technique to investigate a set of multisource remotely sensed data to diagnose plant stress early. The model of this work is the combination of satellite images, high-resolution data using UAVs and a hyperspectral sensor aimed at capturing stress factors, including drought, pests, and nutrient deficiencies, to maximize the health of plants. The system is developed based on state-of-the-art machine learning models, support vector machines (SVM), convolutional neural networks (CNN) and long short-term memory (LSTM) networks are used to accurately analyze and forecast plant stress by utilizing the data. Its efficacy on various datasets is confirmed by experiments that demonstrate impressive results compared to the state-of-the-art approach. The hybrid model is 94.5% higher than the other models, SVM, CNN, and LSTM. In precision agriculture, early warning of the model can be used to efficiently utilize resources, reduce crop losses, and improve yield quality. The given method offers an implementable and scalable solution to the problem that could be applied to commercial-level agricultural systems and stresses the importance of the given method in facilitating the sustainability and agricultural decision-making methods of modern farm systems.
Keywords:	Plant stress, early detection, machine learning, remote sensing, AI, crop health
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	HSCDT: HYBRID SPECTRAL CLUSTERING DECISION TREE FOR PREDICTING CHRONIC KIDNEY DISEASE IN HEART DISEASE PATIENTS
Author:	CHANDRALEKHA E, T R SARAVANAN
Abstract:	Chronic Kidney Disease (CKD) is intertwined with cardiovascular disease (CVD), sharing common risk factors and underlying pathophysiologic processes. However, the early diagnosis of CKD in patients with cardiovascular diseases is crucial for timely intervention and improved patient outcomes. This research introduces a new Hybrid Spectral Clustering Decision Tree (HSCDT) model for predicting CKD in patients with cardiac disease. The method implements spectral clustering and applied decision tree classification for effective discovery of complex patterns in the data set. Various classifiers, including Gradient Boosting, AdaBoost, XGBoost, LightGBM, and Voting Classifier were selected for comparative analysis. The proposed model achieved a peak of 99% cross-validation ac-curacy and a test accuracy of 99%, outperforming previous classifiers. Important clinical predictors of chronic kidney disease (CKD), such as age, body mass index, smoking, history of coronary heart disease, blood pressure, cholesterol, serum creatinine, hypertension, and eGFR, were analyzed and included in the prediction model. The performance on the experimental data confirmed the model’s ability to accurately distinguish the CKD patients from the rest of the population. The findings suggest a significant enhancement in classification effectiveness when spectral clustering and decision trees are consolidated and provide this classification as a possible mode of disease prediction of CKD in the patient suffering from heart disease. Further research can focus on increasing datasets, involving additional biomarkers, and improving models for increased clinical utility.
Keywords:	Chronic Kidney Disease (CKD), Cardiovascular Disease, Hybrid Spectral Clustering Decision Tree (HSCDT), Machine Learning
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	DECIPHERING SUBCLONAL DIVERSITY IN EARLY TUMORS: A NOVEL IMAGING R-NN CYTOMETRY APPROACH TO MAP GENETIC PROFILES AND MARKER EXPRESSIONS
Author:	AYIPUZHA THULASIBAI, BHARATH SINGH JEBARAJ
Abstract:	Tumor heterogeneity is a major issue in cancer diagnosis and therapy due to the difficulty of identifying subclonal variation at an early stage. The goal of this work is to establish a new approach that combines deep learning with high-resolution imaging to identify subclonal heterogeneity in early-stage tumors. Using a Fast R-CNN model to classify subclones and Imaging Mass Cytometry (IMC) to validate the result is a systematic way to detect and visualize the heterogeneity of subclonal populations. The Fast R-CNN had a very high F1-score of 0. 92, which proves the high efficiency of the algorithm in identifying subclonal regions. IMC also provided a strong dual-method validation by associating the genetic markers with the aforementioned subclones computationally. These findings not only extend knowledge of early tumor evolution but also have implications for precision oncology, where therapies can be tailored according to the subclonal architecture. As such, the study shows the possibility of this hybrid approach, but the small sample size and the use of annotated data point to the need for more research. The broader implications of this approach are that this could be applied to other cancer types, which would mean that there is a new approach to the development of multi-cancer detection platforms and personalized treatment. Future work will be directed to the enlargement of datasets, the improvement of models, and the investigation of tumor dynamics over time. The findings of this research provide a solid base for the diagnosis and treatment of cancer, which is a major improvement in the study of oncology.
Keywords:	Subclonal Diversity, Early-Stage Tumors, Fast R-CNN, Imaging Mass Cytometry, Tumor Heterogeneity, Cancer Detection, Personalized Medicine
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	SPR SYNTHESISED: AN EFFECTIVE COLLABORATIVE BASED RECOMMENDATION SYSTEM BASED ON USER SENTIMENTS
Author:	VANTHANA VASAGAN, KARTHEEBAN KAMATCHI
Abstract:	In the current era of electronic and mobile commerce, recommendation system acts as a crucial one in suggesting the products / services to the consumers as well as the users based on their interests. There exist two forms of such systems viz. content based and collaborative based. Among that content-based technique is suitable to suggest the products belonging to the same category. In that extent, collaborative based technique may be a right choice that suits all the sectors. In collaborative techniques, the recommendations may be made on the user based or the item-based similarity. For the similarity calculation, typically the ratings given by the user are used. In some situation, the ratings may be the contradictory or it may not convey the user’s opinion entirely. To prevent this, the proposed study integrates the sentiment analysis with the collaborative-based recommendation systems. Here, along with the rating, polarity of the opinion which indicates the strength of the sentence and the user sentiment is also exploited to compute the similarity score. In such practice, the recommendations given by the system will be more accurate. The results shown that, the accuracy of the system will be more for the item-based filtering when polarity or the synthesis of polarity and rating is utilized for similarity calculation. For the user-based filtering technique, the mixture of polarity, rating and the user sentiment will be the efficient one.
Keywords:	Recommendation System, Sentiment Analysis, Polarity, Collaborative Filtering, User based, Item Based
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	THREAT DETECTION IN FIREWALL LOGS USING FEATURE SELECTION AND ENSEMBLE LEARNING
Author:	NOORDAMA, BENFANO SOEWITO
Abstract:	The firewall serves as the first line of defense in network security, responsible for monitoring and filtering data traffic. However, firewalls remain limited in accurately detecting threats, often generating false positives and false negatives, where legitimate traffic is misclassified as attacks or vice versa. Despite numerous studies that apply machine learning to firewall log analysis, the challenge of accurately distinguishing between benign and malicious activities remains. Existing methods often struggle with high rates of false positives and false negatives, particularly in complex, real-world network environments. This study aims to bridge this gap by combining feature selection and stacking-based ensemble learning, which has been underexplored in addressing the issue of misclassification in firewall logs. Our feature selection approach combines Random Forest feature importance with mutual information, while the ensemble framework employs a stacking strategy that integrates Iterative Dichotomiser 3 (ID3), Random Forest (RF), and Extreme Gradient Boosting (XGBoost), with XGBoost serving as the meta‐learner. Experiments were conducted on two datasets: Palo Alto firewall logs from an Indonesian government agency and a public dataset from the UCI Machine Learning Repository, to evaluate the model’s generalization performance. The results demonstrate that the ensemble learning method reduces false positives and false negatives, achieving 99.97 % accuracy on the primary dataset and 99.87 % on the secondary dataset, representing a 7.14 % reduction in misclassification compared with standalone machine learning algorithms. This work contributes to the development of a robust and efficient threat detection system suitable for deployment in both governmental and public network environments.
Keywords:	Threat Detection, Feature Selection, Ensemble Learning, Random Forest, Network Security.
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	DIFFERENTIAL GENETIC ALGORITHM (DGA) BASED OPTIMAL DIRECTED RANDOM TESTING FOR REDUCING INTERACTIVE FAULTS
Author:	T BALAJI, K PRASUNA, SRINIVAS PADALA, ANJANEYULU NAIK R, 5VENKATA NARAYANA T, G. N. SOWJANYA,
Abstract:	The goal of software testing is to identify software flaws. Software testing is the process of confirming that a program works as intended. Test inputs are generated at random from the software's input space during random testing. to consistently provide test instances that are random and have some similarities. We will provide a method for minimizing the errors based on the best test cases produced by directed random testing in order to get around these problems. Using the object behavior dependence model as a basis, we will create an effective random testing test case in the suggested approach. The Differential Genetic Algorithm (DGA) will be used in this study to generate the best inputs, minimizing both equivalent and illicit inputs. AGA makes advantage of the test case coverage metrics to lessen fault proneness. By merging the old input with the present one, our suggested approach will reduce the input space and improve scalability and efficacy in the software testing age.
Keywords:	Testing, Test Case, Object Behavior Dependence Model, DGA, Coverage Metrics
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	DISTINCTIVE RELATION BASED INTRUSION PREDICTION MODEL USING KERNEL BASED DEEP NEURAL NETWORKS
Author:	YADALA PRABHU KUMAR, BURRA VIJAYA BABU
Abstract:	Implementing Deep Learning (DL) strategies is the latest fad in detecting intrusions in networks. With the proliferation and importance of online interaction, the demand for more robust cyber security measures to stave off cyberattacks has increased. When building a machine learning-based intrusion detection system, attack elements help find, determine, and identify unwanted behavior. The many types of attacks on servers can be detected using behavioral modeling or by building a training model with a learning model. This allows for the identification of network intrusions. This is remedied by employing a supervised learning capability of a Kernel-based Learning Machine (KELM). When compared to more conventional approaches, DL-based intrusion detection can create detection rules with much less data and fewer attack fingerprints. Data-driven learning trains on empirical data and then autonomously defines new characteristics. A high false alarm rate occurs when machine learning models are unable to deal with threats that propagate across applications with a big data footprint or a large number of dimensions. To address these issues, researchers have integrated a Convolution Neural Network into a new kernel-size-based deep learning architecture. The Distinctive Relation based Intrusion Prediction Model is proposed using Kernel based Deep Neural Networks (DRIPM-KDNN) for accurate prediction of intrusions in the network. Following extensive research and simulation on a benchmark dataset, the suggested DRIPM-KDNN model was determined to surpass conventional approaches. The entire network is at risk if even a small number of nodes are compromised. Compared to previous methods, the proposed k-DNN is better since it increases network resilience while decreasing power consumption.
Keywords:	Sensor Network, Link Based Prediction, Topological Analysis, Linkage Identification, Labeling, Generalization, Kernel-Based Deep Neural Networks.
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	A DUAL-HEAD CNN-LSTM FRAMEWORK FOR SIMULATING AND ANALYZING PLANT DISEASE PROGRESSION USING SYNTHETIC TEMPORAL SEQUENCES
Author:	NALLAMOTHU RAGHU, PARDEEP KUMAR
Abstract:	In this study, a novel deep learning framework for plant disease analysis that simulates disease progression using synthetic temporal sequences generated from static leaf images is proposed. Traditional image-based disease classifiers ignore the temporal nature of plant pathology, which limits interpretability and prediction accuracy. To address this, the sequences are synthesised by interpolating between a pseudo-healthy image and a diseased sample using alpha blending. These sequences mimic gradual progression across four frames. The design of a dual-head CNN-LSTM model that takes these sequences as input, leveraging convolutional layers to extract spatial features and LSTMs to capture temporal dynamics. The architecture branches into two heads: one for disease classification via softmax and another for severity estimation through regression. Further integrated a frame-wise attention mechanism and temporal consistency regularisation to improve interpretability and stability. Evaluation on the corn leaf disease dataset demonstrates strong performance, achieving 92.16% classification accuracy and a severity MSE of 0.0001. The comparison of the proposed model against traditional classifiers like Random Forests and Logistic Regression shows a 7–10% improvement in F1-score and significantly better severity awareness. Our approach offers a temporally aware, explainable AI solution for agricultural monitoring. Severity curves and attention maps enable not only post-hoc analysis but also facilitate real-time field-level diagnostics. The pipeline's ability to generate balanced data across underrepresented disease classes is crucial in imbalanced datasets. Moreover, the use of synthetic progression opens opportunities for applying this model to other domains lacking temporal labels. The modular structure of the system supports easy plug-and-play integration with UAVs, mobile apps, and remote sensing platforms. In conclusion, this work bridges the gap between image classification and temporal disease modelling, making a strong case for progression-aware AI in digital agriculture.
Keywords:	Synthetic Temporal Sequences, CNN-LSTM, Plant Disease Progression, Severity Prediction, Frame-Wise Attention, Early Intervention, Dual-Task Learning, Grad-CAM, Corn Leaf Dataset, Interpretable AI.
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	LEVERAGING EFFICIENTNET-B4 IN GOAT DISEASE PREDICTION AND RECOMMENDATION SYSTEMS FOR MORTALITY REDUCTION AND HEALTH OPTIMIZATION
Author:	DR. NILESH B. KORADE, DR. MAHENDRA B. SALUNKE, DR. AMOL A. BHOSLE, DHANASHRI M. JOSHI, GAYATRI G. ASALKAR, DR. PRADNYA A. VIKHAR, DR. PRASHANT B. KUMBHARKAR, DR. SUNIL M. SANGVE, RUPALI UMBARE
Abstract:	Goat farming is a supplementary business to agriculture for Indian farmers, contributing significantly to India's agricultural economy and growth. When a goat gets infected with some disease, the expense of treating a sick goat frequently exceeds the money that could be made from selling the animal. Additionally, there is a risk that the sickness could infect other goats, leading farmers to isolate the sick goat from the rest of the flock. Due to the low survival rate and reduced weight of the goat after recovery, farmers usually concentrate less on treating the infected goat, as it is unlikely to be profitable even if it survives. In this research, we build a deep learning-based framework to address the problem of early disease identification in goats. A dataset of 1960 images, representing six major goat diseases, was prepared by visiting different goat farms. The preliminary preprocessing techniques, including image resizing, normalization, and noise reduction, were applied to the collected dataset of goat disease images in order to enhance the quality and consistency to boost model training and performance. Several architectures, including CNN, AlexNet, VGG-19, ResNet-50, and EfficientNet-B4, were trained on gathered data and evaluated using evaluation measures. Among all the architectures evaluated, EfficientNet-B4 achieved an excellent accuracy of 93%, demonstrating its robustness and efficiency in diagnosing goat sickness. In comparison, CNN achieved 72%, AlexNet delivered 79%, VGG-19 delivered 77%, and ResNet-50 delivered 89% accuracy. The proposed framework demonstrates strong potential as a feasible option for real-time goat health monitoring, offering farmers a useful tool for assisting with early detection and prevention. This advancement can accelerate efforts toward achieving sustainable and profitable livestock production in rural India.
Keywords:	Goat Farming, Disease Prediction, EfficientNet-B4, Livestock, Health.
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	DIGITAL TECHNOLOGIES AS A TOOL FOR INCREASING THE BUSINESS COMPETITIVENESS
Author:	LIUDMYLA VERBIVSKA , OKSANA ZYBAREVA , KATERYNA OZARKO , OLHA KIRICHENKO, OKSANA REDKVA
Abstract:	Digital technologies provide businesses with numerous benefits, but their positive effect of enhancing competitiveness is most strongly manifested in interaction with other influencing factors, in particular, innovations. The aim of the study is to assess the joint impact of digitalization and innovation on the business competitiveness taking into account macroeconomic and corporate factors. The study employed the methods of regression, variance, and correlation analysis. The analysis found that digitalization has a statistically significant impact on the competitiveness of countries through the indicators Digital Skills Among Active Population and Individuals Using the Internet. The relative impact of these indicators on the competitiveness of countries is 0.6833 and 0.3024, respectively. Innovation plays a mediating role in the impact of digitalization on competitiveness through the Institutions indicator. This gives grounds to conclude that enhancing the competitiveness of the country and the business sector requires not only technological development. It also requires improving the quality of the institutional environment to fully realize the potential of digital technologies. Analysis of the impact of companies’ research and development (R&D) spending on revenue revealed a positive correlation between these indicators. With an increase in R&D spending by 1 unit, expected revenue increases by 0.5571 units. Accordingly, an increase in R&D spending has the potential to increase the efficiency of business activities, which ultimately leads to increased revenue through innovation and process optimization. The findings can be useful for entrepreneurs for improving business strategies by focusing on the identified determinants of the impact of digitalization on competitiveness. Determining the role of innovation in this process is also important for innovation policymakers to optimize the environment for enhancing competitiveness through the effective use of digitalization and innovation.
Keywords:	Digitalization; Innovation; Institutions; Competitive Advantage; Entrepreneurship
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	MULTI-LEVEL FEATURE EXTRACTION AND SELECTION WITH STRONG CORRELATION AND EFFECTIVE CLUSTERING MODEL FOR ACCURATE MELANOMA DETECTION
Author:	ZAHEER ABBAS MOHAMMAD, S.NAGAKISHORE BHAVANAM, VASUJADEVI MIDASALA
Abstract:	Melanoma, a fatal and aggressive form of skin cancer has continuously increased in incidence rates across the globe, emphasizing the necessity for early precise detection modalities to achieve better 5 year survival outcomes. This research presents a significant contribution to the field of Information Technology by proposing an advanced automated framework for accurate melanoma detection through a combination of multi-level feature extraction, strong correlation analysis, and intelligent clustering. With the ever-increasing volume of medical imaging data and the limitations of manual diagnosis, IT-based solutions such as machine learning and image processing are not just supportive tools they are transformative technologies driving modern healthcare innovation. This work underlines the indispensable role of IT in developing intelligent, data-driven healthcare systems capable of early diagnosis, ultimately contributing to increased survival rates and reduced healthcare burdens. This research proposes a Multi Level Feature Extraction and Rank Linked Feature Set with Clustering (MLFE-RLFSC) Model for accurate melanoma detection. Furthermore, this intelligent feature selection process helps in reducing the risk of over fitting and also improves significantly generalization performance of model to classify melanoma more accurately. In addition, our method integrates a clustering model which groups the type identical features so overall redundancy can be diminished as well that in turn enhance the general security level of detection systems. The proposed model achieved 98.8% accuracy in Multi Level Feature Extraction, 99.2% accuracy in Feature Rank Allocation, 98.6% accuracy in Clustering, 99.3% accuracy in Rank Linked Feature Set Generation and 99.2% accuracy in Classification of benign and malignant tumours. The proposed model when compared to the traditional methods performs better in clustering and melanoma detection. The proposed model with minimum false predictions accurately detects the melanoma for providing early diagnosis.
Keywords:	Melanoma Detection, Multi-Level Feature Extraction, Rank Linked Feature Set, Clustering Model, Image Processing, Melanoma Cancer Diagnosis.
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	A HYBRID SVD-DIRECTIONAL FILTERING APPROACH FOR HIGH-DENSITY SALT-AND-PEPPER NOISE REMOVAL FROM GRAYSCALE IMAGES
Author:	ALI SALEM AL RAWASH, FARAH AINI ABDULLAH, AHMAD KADRI JUNOH, ACHMAD ABDURRAZZAQ
Abstract:	One of the fundamental challenges in image processing is denoising grayscale images affected by high-density salt-and-pepper noise, where existing median, adaptive, and wavelet-based filters often fail to preserve fine details and structure at noise levels above 70%. To address this gap, this study presents a novel hybrid denoising approach that combines directional filtering with singular value decomposition (SVD). The proposed method leverages directional filtering to preserve edge and textural features while SVD-based reconstruction reduces residual noise through adaptive low-rank approximation. This combination introduces a robust solution for grayscale image denoising at extreme noise densities. Experimental results demonstrate significant improvements in PSNR and SSIM over conventional techniques, providing new knowledge on the effectiveness of hybrid filtering-decomposition methods for image restoration tasks, particularly in scenarios where edge preservation and structural integrity are critical.
Keywords:	Salt and Pepper Noise, Image Denoising, Image restoration, Singular Value Decomposition, Directional Filter
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	THE ROLE OF SOCIAL MEDIA ANALYTICS IN GOVERNMENT DECISION-MAKING: A MACHINE LEARNING APPROACH USING COVID-19 VACCINATION TWEETS
Author:	Social media has emerged as a vital platform for expressing public opinion and disseminating information. During the COVID-19 pandemic, governments worldwide leveraged social media platforms to inform citizens about policies, preventive measures, and vaccination campaigns. Analyzing these vast amounts of data using machine learning and social media analytics (SMA) provides crucial insights to support policy development and strategic decision-making. This study explores the role of SMA in government decision-making during the COVID-19 pandemic by evaluating sentiment classification on English-language tweets related to vaccination. We created a novel COVID-19 vaccine-related tweet dataset and applied six machine learning algorithms—Random Forest (RF), Support Vector Machine (SVM), XGBoost, Decision Tree (DT), Naïve Bayes (NB), and Logistic Regression (LR)—to classify tweet sentiments. Performance metrics, including accuracy, precision, recall, and F1-score were used to compare model performance. The results indicate that the Random Forest model outperformed others, achieving the highest accuracy, precision, recall, and F1-score. This demonstrates the potential of SMA combined with machine learning in capturing public sentiment and supporting informed government responses during public health crises.
Abstract:	Social Media Analytics, COVID-19 Vaccination, Sentiment Analysis, Machine Learning, Government Decision-Making
Keywords:
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	OPTIMIZING CLOUD COSTS AND CARBON FOOTPRINT IN MULTI-CLOUD ENVIRONMENTS WITH FUZZY LOGIC & MONTE CARLO SIMULATION
Author:	UDAY KUMAR DOSANAPUDI, SUNEETA MOHANTY, PRASANT KUMAR PATTNAIK
Abstract:	Organizations are using multi-cloud architectures increasingly to address performance and scalability needs in a world where data is king. But higher operational expenses and carbon emissions are regular results, which is bad for the environment. Using fuzzy logic and Monte Carlo simulations, this study presents a novel hybrid optimization method for lowering the cost and carbon footprint of cloud computing. The suggested solution changes the way cloud resources are allocated in real time based on things like workload, pricing, and the weather. Fuzzy logic may help you make decisions when the inputs are not clear, and Monte Carlo simulations can help you figure out how much effort and how much money you need. When we compare our technique to static baseline approaches, testing using real-world datasets demonstrate that our approach cuts operational expenses by 25% and carbon emissions by 33%. These findings suggest that there is space for clever, flexible ways to manage resources in a manner that is good for the environment across several clouds.
Keywords:	Cloud Cost Optimization, Carbon Footprint Reduction, Fuzzy Logic, Monte Carlo Simulation, Multi-Cloud Environment
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	BLENDED LEARNING AND EMERGING TECHNOLOGIES IN MATHEMATICS EDUCATION: A COMPREHENSIVE STRUCTURED REVIEW
Author:	NURSAADAH JAILANI, ROSLINDA ROSLI, MUHAMMAD SOFWAN MAHMUD
Abstract:	Studies on blended learning and emerging technologies have transformed mathematics teaching and learning through innovative educational approaches. However, integrating these methods presents challenges, including technological hardships, instructional design complexities, and the need for appropriate digital tools. This study conducted a systematic literature review to analyse the features, issues, and prospects of implementing new technologies in mathematics education using thematic analysis. An empirical review was conducted using Scopus and Web of Science (WoS) databases, selecting scholarly publications from 2023 to 2024. The study followed the PRISMA framework, identifying 33 primary studies, which were categorized into three themes using data analysis: (1) blended learning and flipped classroom models, (2) virtual and embodied learning environments, and (3) teacher perspectives, professional development, and student learning experiences. Findings indicate that blended learning enhance student interest, attitude, and achievement, yet their effectiveness critically depends on teachers' technological and pedagogical content knowledge (TPACK) and institutional support. Additionally, the increasing use of artificial intelligence (AI) in mathematics education presents opportunities but also challenges, including implementation limitations, expertise gaps, and funding constraints. This study highlights the critical role of professional development in preparing educators to support blended learning effectively. Several recommendations are proposed for stakeholders to overcome challenges and maximize the benefits of emerging technologies. Findings provide valuable insights for researchers, educators, and policymakers in enhancing mathematics education to meet 21st-century learning demands.
Keywords:	Blended Learning, Digital Tools, Mathematics Technologies, Mathematics Teaching Innovations, Teacher Technology Proficiency.
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	ENHANCED DENSE CONVOLUTIONAL NETWORK FOR BRAIN TUMOR CLASSIFICATION USING MODIFIED BIRCH SEGMENTATION
Author:	CHIRANJEEVI K, VICTO SUDHA GEORGE G
Abstract:	The ability and experience of a radiologist is essential for the long and challenging process for detecting brain tumors. With the increasing number of patients, the number of data is maximized, making the fastest processes more expensive and inefficient. Many researchers have looked at many faster, more accurate methods to identify brain tumor classifications. In particular, Deep Learning techniques are detected to create automated systems that can achieve proper identification or segmentation of brain tumors in a shorter period of time. This article suggests a classification of four stages of brain tumors based on Edenet. The specified images are processed in the first stage using GF, and the pre-processed images are segmented using modified birch segmentation in the second stage. After segmentation, features such as shape functions (moments, area, scopes, epsilons, convex) are extracted as a third stage as new variants of texture functions (improvements in LGTP), MBP, and residual functions. In the final phase of the specified model, Edenet-based classification is performed for the classification of brain tumors, and its performance is verified using traditional methods in terms of various performance measurements.
Keywords:	Brain Tumor,Dl, Birch, Edensenet,Iltp
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	AI-POWERED INTRUSION RESPONSE FOR INTELLIGENT VEHICULAR ECOSYSTEMS
Author:	L. K. SURESH KUMAR, RAVI UYYALA, JAIDEEP GERA, A L SREENIVASULU, P. SASIREKHA, KUNCHANAPALLI RAMA KRISHNA
Abstract:	Implementing strict cybersecurity measures to protect against cyber-attacks is absolutely necessary, given the increasing number of intelligent vehicles on the road. This study aims to learn more about the potential for creating an intelligent vehicle-specific autonomous intrusion response system (IRS). The proposed IRS system can instantly assess the consequences of intrusions and ascertain the best methods of response depending on the situation. Among the most significant contributions are a thorough analysis of different response techniques, a system for evaluating costs and impacts dynamically, and the application of various selection algorithms including Simple Additive Weighting (SAW), Linear Programming (LP), game theory, and AI-based procedures. Research has shown that the system works well in terms of response quality, efficiency of time, and consumption of resources. This proves that the technology has the ability to greatly enhance car safety. The findings of this study lay the groundwork for future framework improvements and adaptations by the Internal Revenue Service.
Keywords:	Intrusion response system, Cybersecurity, Intelligent vehicles, Linear Programming, Game theory, AI-based mechanisms
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	OPTIMIZED LSTM-DRIVEN MRI PREPROCESSING WITH FINE-TUNED U-NET FOR ACCURATE BRAIN TUMOR SEGMENTATION
Author:	KETHA. SHANTHI, DR. S. RADHA KRISHNA
Abstract:	The evaluation of brain tumors within MRI scans stands as a fundamental medical imaging operation which benefits diagnostic procedures as well as treatment arrangement processes. The research investigates an optimized arrangement of Long Short-Term Memory (LSTM) networks that process MRI scans before using a modified U-Net network for brain tumor segmentation. The sequential LSTM model processes MRI sequence images so it improves spatial coherence and reduces noise to enhance image quality by adapting intensity normalization. The initial processing operation enhances the quality of the input data thereby improving its readiness for segmentation. A U-Net architecture with fine-tuned capabilities uses EfficientNet as its backbone in the encoder section while adding attention mechanics to the decoder framework. The model optimization strategy includes the combination of Dice and Focal losses for effective class imbalance management. The model performance benefits from additional improvements which come from both data augmentation methods and meticulous hyperparameter adjustments. Research experiments performed on standard MRI datasets show that the proposed method outperforms past methodologies at achieving higher segmentation outcomes. The proposed method attaining 10.5%, 7.11%, 1.26 improvement with reference to Dice Similarity Coefficient (DSC), 8.56, 6.91, 2.36 improvement with reference to Jaccard Index (JI) and 10.23, 5.69, 3.94 improvement in segmentation accuracy over the U-Net, Attention U- Net, Res Attention U-Net models. The advanced U-Net architecture together with sequential LSTM preprocessing allows the system to produce both effective feature detection and accurate tumor outline identification. The approach demonstrates prospective capability to support radiologists in their automated diagnosis of tumors and therapeutic planning activities. The authors will analyze real-time implementation and clinical application optimization during subsequent stages of their research.
Keywords:	Brain Tumor, Fine Tuned U-Net, LSTM, MRI Image, Pre-processing, Segmentation.
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	REMOTE SENSING IMAGES FOR WATER QUALITY MONITORING BASED ON DEEP LEARNING MODEL
Author:	SOFIA PRIYADHARSHINI D , G P RAMESH
Abstract:	For the growing number of ecological issues that adversely affect natural systems, accurate assessment and monitoring of ecological functioning was essential as a prerequisite for a better response. To obtain a reliable performance of environmental control and the following evaluation, this study built an integrated innovative framework with 3 layers that combine CMT, DL and RST. To acquire better precise and enhanced monitoring results, the FMRCNN a type of DL modifies and efficiently utilizes the big data platform offered by RST and CMT. Our case study at PIL showed that this proposed system has an efficient performance than the existing system. The time series were created for each feature in PIL water bodies using all available data, covering the years 2017 to 2022. Next, we find unusual data points on time series using the EDA anomaly detection approach. The EDA approach successfully recognises abnormal events in WQ based on other studies. Furthermore, the relationship between non-optically active variables such as pH, DO, BGA, Chl, Turbidity and fDOM is very strong, according to our findings. The proposed methodology improves the performance measures reliable and effective than the existing system to monitor indoor WQ at the PIL.
Keywords:	Data Analytics; Deep Neural Network; Machine Learning; Regression; Water Quality
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	THE INSPECTION OF VARIOUS INPUT TRANSFORMATIONS TOWARDS HUMAN DETECTION AND TRACKING
Author:	TANVEER KADER, AHMAD FAKHRI AB. NASIR , ANWAR P.P. ABDUL MAJEED, M. ZULFAHMI TOH, NOORLIN MOHD ALI
Abstract:	Considering the role of color space in enhancing object detection this paper finds the impact of detection models trained on different color spaces in human tracking system within the Tracking by detection (TBD) framework. A customized dataset over 8k frames including indoor and outdoor human movement videos was developed following the MOT15 structure. YOLO12s was fine-tuned separately on six color spaces RGB, Grayscale, HSV, HSI, YCbCr and YES followed by SORT and DeepSORT for tracking. YOLO12s trained on RGB provides the best MOTA with 31.5% and 50% for SORT and DeepSORT respectively. Competitive results observed from the Grayscale color space.
Keywords:	Human tracking, tracking-by-detection, color spaces, SORT and DeepSORT
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	A SMART ANTI-PHISHING MODEL FOR PHISHING WEBSITE DETECTION USING MACHINE LEARNING APPROACHES BASED ON HYBRID FEATURES
Author:	M.VENKATA KRISHNA REDDY, S.CHINA RAMU1, P.RAMESH BABU, P.NIRUPAMA, B.RAMAKANTHA REDDY, KADIYALA RAMANA
Abstract:	Phishing is a major concern in a changing society. The rise of the Internet has led to a new form of data theft, known as cybercrime. One of the most prevalent forms of cybercrime, phishing attempts to trick users into divulging personal information by creating a convincing look and feel of a trusted online service, like a bank, grocery store, or online media website. The problem of detecting phishing websites has been discussed on multiple platforms, with approaches varying from straightforward classifiers to complex hybrid systems. A novel phishing detection system, “Phishing URL Detection, PUD”, is proposed here. It uses machine learning approaches to analyze results from various methods applied to URLs and validates them against existing research. URL-based phishing is a prevalent method to collect user data when accessing a malicious website. Detecting rogue URLs is difficult. The proposed method seeks to discover such websites using machine-learning approaches that analyze the behavior and attributes of the suggested URL. To better understand the structure of malicious URLs, various machine-learning methods were tested for feature evaluation. Precise parameter tuning facilitates choosing the best machine learning method for identifying malicious from legitimate websites. One of the major goals is to train machine learning models to find and prevent phishing websites using the dataset. Various models' levels of performance are evaluated and contrasted. The proposed system outperforms state-of-the-art models and demonstrates the importance of hybrid URL features in phishing website detection.
Keywords:	Cybercrime, Phishing, Legitimate websites, URLs, Machine learning approaches, Detection
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	A SYSTEMATIC REVIEW: EMPLOYING AI IN ADAPTIVE LEARNING RECOMMENDATION SYSTEM FOR VOCATIONAL EDUCATION
Author:	FATMA A. SETYANINGSIH1, HERMAN D. SURJONO, SRI ANDAYANI
Abstract:	The integration of AI in vocational education, especially in adaptive learning, highlights the importance of automatically detecting individual learning styles. Traditional methods such as questionnaires, though reliable, face limitations like student reluctance and lack of self-awareness. This reveals a research gap in learning style detection, particularly in AI-based adaptive systems, requiring further exploration of effective computational techniques in real-world educational contexts. These challenges are especially relevant to the development of adaptive learning-based recommendation systems for career selection. Accurate learning style detection is essential not only for personalized learning but also for aligning educational content with potential career paths, thereby enhancing both academic and career outcomes according to talent, interests, and major. To address these gaps, this study presents a systematic review of 40 selected articles published between 2014 and 2025. The review examines techniques, approaches, and computational strategies used in automatic learning style detection and their implementation in various vocational educational settings. Findings show that AI, particularly data-driven approaches, significantly supports learning adaptation. The Felder–Silverman model and classification techniques like K-Means and Naive Bayesian are commonly applied due to their adaptability across contexts. Moodle also emerges as a frequently used platform for data collection and experimentation. These insights are fundamental for designing intelligent recommendation systems that adapt to student’s learning styles and support personalized career guidance. Integrating such systems can enhance educational relevance, improve learning outcomes, and foster long-term career readiness.
Keywords:	Artificial Intelligence, Career Recommendation, Vocational Education, Adaptive Learning.
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	IMPACT OF MULTI-PHASE, MULTI-FREQUENCY CLOCKING ON DELAY AND SUPPLY NOISE IN MODERN CMOS PROCESSORS
Author:	R. KANNAN, M. JENISH, LIZATH SAHIR, SALITH SAHIR, K.NAGARAJAN, L.MOHANA KANNAN
Abstract:	The supply noise and latency of modern CMOS processors are investigated in this work to see how multi-phase, multi-frequency clocking influences these characteristics. The increasing demand for improved performance and energy efficiency in semiconductor devices has led to the development of optimal clocking algorithms, which are becoming increasingly important components. In the beginning, research studied about how CMOS processors are now having a hard time meeting the ever-increasing needs for performance while consuming a very little amount of power. The number of research that have been conducted in the past on the impacts of delay and supply noise on single-phase, multi-phase, and multi-frequency clocking operations has been rather limited. Our proposed method makes use of cutting-edge CMOS process models and advanced simulation tools to carry out a comprehensive study. This is done to bridge the gap that has been identified. To determine the effects that different clocking configurations have on delay and supply noise across a spectrum of frequencies and phase relationships, these configurations are subjected to stringent testing. The findings shed insight on the costs and benefits of multi-phase, multi-frequency clocking, which were previously unknown. During our research, research discovered that certain configurations drastically cut down on supply noise while simultaneously raising latency. This could be an effective method for improving CPU performance in general. In addition to this, the study identifies specific instances in which the associated costs and benefits need to be carefully considered.
Keywords:	Clocking, Multi-frequency, Multi-phase, CMOS Processors, Delay, Supply Noise
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	COMPANY INFORMATION SECURITY MANAGEMENT MECHANISMS FOR CRITICAL DATA PROTECTION
Author:	YEVHENII IPPOLITOV, IVAN KYDRIAVSKYI, ALLA BALAN, DINA DRYZHAKOVA, IVAN IURIEV
Abstract:	The relevance of this study is determined by the growing need to protect company’s critical data from cyber threats, which is an important part of the information security strategy. The aim of the study is to assess the level of implementation of information security mechanisms among companies and determine the effectiveness of their application for data protection. The research employed the following methods: questionnaire surveys, comparative data analysis, and correlation analysis. The obtained results showed that the lack of regular monitoring (0) correlates with a high frequency of incidents (3.9, correlation -0.97), which confirms the insufficient cyber protection. The frequency of incidents has an inverse relationship with the level of technology implementation (-0.96), especially in construction, where modern solutions are almost not used. The level of response to incidents is positively correlated with the effectiveness of protection mechanisms (0.99), which proves the importance of a quick response to threats. The academic novelty of the study is the comparison of the levels of implementation of security policies in different sectors and determining the impact of the effectiveness of company information security management mechanisms on ensuring data security in the face of modern cyber threats. Research prospects include a deeper study of the impact of external and internal factors on the effectiveness of information security.
Keywords:	information security, cyber threats, outsourcing, security policy, data protection, IT sector.
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	DEEP LEARNING ARCHITECTURES FOR EXPLORING SPATIO-TEMPORAL PATTERNS FROM EEG DATA FOR EMOTION DETECTION
Author:	ANDHAVARAPU BHANUSR, PROF. MOGALLA SHASHI
Abstract:	Human emotions are crucial across various domains by closing the divide between human and current technologies, thus fostering better understanding on critical mental health conditions. Traditional approaches rely mainly on facial expressions along with body gestures and are met with limited success as they miss underlying or repressed emotions. Electro Encephalo Gram (EEG) signals offer a direct glimpse into brain activity, making them a promising avenue for fool-proof emotion recognition. However, the complex temporal dynamics of EEG data pose challenges for classical machine learning algorithms, which often fail to capture spatial and temporal patterns effectively. To overcome these issues, we introduced a novel model called Spatio-Temporal Difference Identification - Convolution Neural Network (STDI-CNN). Our model efficiently captures the complex temporal dynamics of EEG data, indicating brain activity across time in various lobes, by utilizing deep learning approaches, especially a combination of CNN and sequential neural network architectures. Extensive experiments on the SEED EEG dataset demonstrate the efficacy of the proposed STDI-CNN model, achieving an impressive accuracy of 98.52%. Additional tests using CNN-LSTM and CNN-BiLSTM models also yielded strong performance, with accuracy rates of 97.04%. This surpasses current SOA models and highlights the potential of STDI-CNN in extracting meaningful patterns from EEG signals for emotion recognition. Our work reduces the gap by featuring a significant step forward in harnessing EEG signals to build well informed emotionally intelligent system that fosters prior detection and improved diagnosis for neurological disorders.
Keywords:	EEG Emotion Recognition, CNN For Emotion Recognition, Fusion of CNN and LSTM, Spatial and Temporal Patterns, SEED Dataset.
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	BLOCKCHAIN-BASED EDGE COMPUTING: JOINT TASK OFFLOADING AND MINING WITH MULTI-AGENT REINFORCEMENT LEARNING
Author:	R. ANIL KUMAR, SARALA PATCHALA, SUNITHA PACHALA, GEETA BHIMRAO ATKAR, U.H.B.K. MAHALAXMI, ANGARA SATYAM
Abstract:	The integration of mobile edge computing (MEC) and blockchain enhances computing efficiency and security. This paper presents a novel cooperative task offloading and block mining (TOBM) approach in blockchain-based MEC. The system enables edge devices to offload tasks and participate in mining operations. To handle blockchain latency, a novel Proof-of-Reputation (PoR) consensus mechanism is introduced. A multi-objective function is developed to optimize system utility by managing offloading, channel selection, power allocation and computational resources. A multi-agent deep deterministic policy gradient (MA-DDPG) algorithm is used for optimization. A game-theoretic approach is applied to model competition among edge devices and establish a Nash equilibrium. Simulations demonstrate improved system utility compared to traditional approaches. The proposed TOBM framework provides efficient task allocation and computational resource management. It dynamically adapts to network conditions, reducing computational delays and enhancing overall performance. Blockchain security mechanisms prevent unauthorized data modifications, promoting data integrity and trustworthiness. The PoR consensus mechanism minimizes the verification time required for block mining, allowing for faster transactions and reduced network congestion. The proposed method enables edge devices to make intelligent task offloading decisions, optimizing computational efficiency while maintaining low energy consumption. The MA-DDPG algorithm effectively learns from network interactions, continuously improving decision-making policies. The results indicate that the system significantly outperforms existing solutions in terms of latency reduction, resource utilization and security enhancements. Future research directions include enhancing the PoR mechanism and exploring additional consensus models to improve scalability and performance.
Keywords:	MEC, TOBM, Task Offloading, Blockchain Design, MADRL, MADDPG.
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	FINE-TUNING CHEMBERTA TRANSFORMER MODEL FOR SUPERIOR MOLECULAR PROPERTY PREDICTION (MPP)
Author:	Dr.A. SANDHYA, Prof. CHIN-SHIUH SHIEH, Prof. MONG-FONG HORNG, Prof. R. SUBHASHINI, Dr. R. SETHURAMAN
Abstract:	Drug discovery is expensive and time-consuming, often taking decades and billions of dollars for its research and development to analyze its chemical properties for drug development. The task of Molecular Property Prediction (MPP) is major crucial activity for selecting potential drug candidates in drug Discovery. The chemical language models (CLMs) have gained attention in this area due to their success in natural language processing tasks like text analysis and translation. The availability of vast unlabeled chemical data has driven the development of CLMs to interpret molecular information effectively. This study uses a large language model approach by fine-tuning a Transformer model to predict molecular properties such as solubility, toxicity and pIC50. Traditionally, molecular graphs and descriptors were used for property prediction. However, Transformer-based models like BERT, GPT, and their variations have shown exceptional performance in downstream NLP tasks. In this research, the chemical informatics, model such as ChemBERTa have been explored for learning molecular contextual information from Simplified Molecular Input Line Entry System (SMILES) sequences. Experimental results show that ChemBERTa outperforms traditional ML models in molecular property prediction, achieving Accuracy and AuROC scores of 0.94 and 0.96, respectively. This demonstrates its superior predictive capability compared to state-of-the-art methods. The study highlights the potential of Transformer-based CLMs in accelerating drug discovery by effectively predicting molecular properties from text-based inputs like SMILES strings.
Keywords:	Natural Language Processing (NLP), Transformer Model, Pre-Training, Fine Tuning, chemBERTa, SMILES, Cheminformatics, Molecular Property Prediction (MPP).
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	ENHANCED ERROR CORRECTION FOR MEMORY CELLS IN SPACE: A DUAL APPROACH TO RANDOM AND BURST FAULT DIAGNOSIS
Author:	P V GOPIKUMAR, R MANIKANDAN, C RAVI SHANKAR REDDY
Abstract:	This paper presents a novel method of fault diagnosis in memory cells that can be employed for space applications. The method presented here capable of diagnosing both random faults as well as burst errors. The proposed method is composed of two techniques one of the technique aims at targeting the random faults while the other aims at diagnosing the burst errors. The diagnosis of random faults is obtained by employing the extended decimal matrix coding whereas the burst errors diagnosis is done by Flexible Unequal Error Method. This method gains an adequate advantage in increasing the length of burst errors coverage based on the probability of distribution of the random faults. In addition to these little overhead bits were added to increase the diagnosis of burst errors. Experimental results conclude that the proposed error correction code achieves 100% of error correction rate for 5bit adjacent errors and six-bit random errors and better error correction rates up to 8-bits compared other ECC’s that are considered for experimental purpose. However, this enhancement in error correction rate is achieved at a cost of slight increment in overhead of redundant bits.
Keywords:	Random errors, Burst errors, Encoder, Decoder, Modified Decimal Matrix code, Flexible Unequal Error control (FUEC).
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	THE IMPACT OF GENDER MODERATION ON AUDITORS’ INTENTION TO ADOPT CYBERSECURITY: A TAM AND PMT FRAMEWORKS APPROACH
Author:	BAGAS INDRA PRANATA, FIDELA ANDINA, IGNATIUS EDWARD RIANTONO
Abstract:	The growing complexity of cyber threats in the auditing environment has intensified the need to understand the factors that drive auditors to adopt cybersecurity technologies. This study explores the impact of gender moderation on auditors' intention to adopt cybersecurity technologies by utilizing the Technology Acceptance Model (TAM) and Protection Motivation Theory (PMT). A quantitative approach was employed, with data collected from 429 auditors working in public accounting firms across Indonesia through structured questionnaires. After excluding 16 outliers with highly fluctuating responses, the final sample consisted of 413 respondents. Structural Equation Modeling-Partial Least Squares method (PLS-SEM) was employed using SmartPLS to analyze the data. The findings indicate that Perceived Usefulness, Perceived Severity, Perceived Vulnerability, Perceived Response Efficacy, and Perceived Self-Efficacy significantly influence auditors' intention to adopt cybersecurity technologies. However, Perceived Ease of Use does not have a significant impact. Additionally, gender did not moderate the relationship between these factors and auditors’ adoption intention. These results suggest that although perceptual differences between male and female auditors exist, they are not substantial enough to warrant gender-specific cybersecurity adoption strategies. Therefore, cybersecurity implementation policies can be applied universally across auditors irrespective of gender.
Keywords:	Cybersecurity, Audit Technology, Data Security, Technology Acceptance Model, Protection Motivation Theory
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text

Title:	TRUXL: TRUST-BASED SECURE ROUTING AGAINST RPL ATTACKS IN IOT USING XGBOOST WITH CNN-LSTM
Author:	R. ELANGO, DR. D. MARUTHANAYAGAM
Abstract:	One of the most popular routing protocols in Internet of Things (IoT) contexts is the Routing Protocol for Low-Power and Lossy Networks (RPL). However, RPL is extremely vulnerable to a number of security risks that jeopardize data integrity and network dependability, including as wormhole, Sybil, blackhole, and rank assaults. This research suggests TRUXL (Trust-Based Routing using XGBoost and CNN-LSTM), a hybrid machine learning (ML) and deep learning (DL) architecture for trust-aware secure routing in Internet of Things networks, as a solution to these problems. The TRUXL model combines a CNN-LSTM hybrid technique for dynamic trust score prediction with XGBoost for trust-based node classification. While CNN extracts geographical trust patterns and LSTM captures temporal trust changes for improved anomaly detection and attack mitigation, XGBoost efficiently classifies IoT nodes based on energy levels, packet forwarding behavior, and historical anomalies. Real-time trust evaluation is used to develop a secure routing technique that mitigates security vulnerabilities by dynamically choosing the most dependable paths. The suggested model's performance is compared to that of DBN-TSP (Deep Belief Network with Trust Score Propagation) and RF-Trust (Random Forest-Based Trust Model) utilizing OMNeT++ with the INET Framework. Packet Delivery Ratio (PDR), End-to-End Delay, Routing Overhead, Anomaly Detection Accuracy, Trust Score Stability, and Attack Mitigation Rate are among the evaluation measures. According to experimental data, TRUXL performs better than conventional trust-based routing models in RPL-based IoT networks, achieving enhanced routing efficiency, adaptive security, and higher attack detection accuracy. The suggested TRUXL offers a clever, scalable, and reliable solution for secure IoT routing by combining CNN-LSTM for trust prediction with XGBoost for classification. This greatly improves trust-aware communication in resource-constrained contexts.
Keywords:	Internet of Things (IoT), RPL Security, Trust-Based Secure Routing, XGBoost, Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), Machine Learning (ML), Deep Learning (DL) and Anomaly Detection.
Source:	Journal of Theoretical and Applied Information Technology 31^st August 2025 -- Vol. 103. No. 16-- 2025
Full Text