		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
November 2025 | Vol. 103 No.21

Title:	YOLO-BASED FEATURE LEARNING WITH RF–XGB ENSEMBLES FOR ROBUST CITRUS LEAF DISEASE DETECTION
Author:	AJAY KUMAR VEGI, Dr.PADMAJA.PULICHERLA, Mr.SANJIT KUMAR BARIK, Dr. P THIRUMOORTHY, Dr. VIJAY DHAWALE, Mr. SITANSHU KAR, Mr.MIHIRKUMAR B. SUTHAR, Dr. SIVA KUMAR PATHURI
Abstract:	Diseases of citrus leaves regularly harm fruit production worldwide, causing big financial losses and lower yields. Understanding a patient’s condition and how severe it right away helps a lot in successful treatment. The work shows how to automatically diagnose and rate citrus leaf disease by applying deep learning and machine learning ensemble methods. Using the YOLO algorithm, the suggested approach can spot and identify parts of the leaves that look unhealthy in real time. Following identification, a hybrid ensemble classifier which combines RF and XGBoost is used to assess the importance of each area and how severe it is. Teaching and testing of the system were possible using pictures of citrus leaves that show several different diseases and levels of severity. Even though YOLO adapted illness areas well, the ensemble consisting of RF-XGB came out on top in the severity task by dealing smoothly with both complex links between features and random noise. The model showed its effectiveness in detection and classification since it produced high scores for accuracy, precision, recall, and F1-score. The combination helps farmers and agriculture experts swiftly and correctly monitor their crops in citrus gardens. The use of this digital model may benefit smart agriculture, mobile medical diagnoses, and supervising wide-ranging plant diseases easily. The proposed classifier gave the best accuracy of 96% which when compared with base classifiers like LR and Decision Tree. ‘
Keywords:	Deep Learning, LR, Citrus Leaf, Decision Tree, YOLO+RFXGB(Proposed).
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	ENHANCED DISTRIBUTED ADAPTIVE OPPORTUNISTIC ROUTING IN MANETS USING KRILL HERD OPTIMIZATION (EDAOR-KHO)
Author:	P.BALAMURUGAN, DR.S.DHANALAKSHMI
Abstract:	Mobile ad hoc networks (MANETs) are characterized by dynamic topologies, limited resources, and decentralized routing challenges. Opportunistic routing, which relies on multiple potential forwarders instead of predefined paths, has emerged as a promising approach to enhance packet delivery. To address persistent issues such as high latency, network congestion, and energy inefficiency, this study proposes an Enhanced Distributed Adaptive Opportunistic Routing method integrated with the Krill Herd Optimization algorithm (EDAOR-KHO). The proposed approach dynamically selects optimal forwarding nodes by considering multiple performance factors, including transmission reliability, network lifetime, energy consumption, and packet delivery ratio. Comparative analysis with established protocols—EXOR, d-AdaptOR, EDAOR-AFS, EDAOR-FF, and EDAOR-GWO—demonstrates that EDAOR-KHO consistently outperforms existing techniques. Simulation results confirm that EDAOR-KHO reduces end-to-end delay, lowers routing overhead, improves packet delivery reliability, and extends overall network lifespan.
Keywords:	Mobile Ad-Hoc Networks (MANETs), Opportunistic Routing, Distributed Adaptive Opportunistic Routing (d-AdaptOR), Krill Herd Optimization (KHO), Routing Optimization,Packet Delivery Ratio (PDR),Energy Efficiency,Network Lifetime,Swarm Intelligence, Wireless Communication
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	A NOVEL INTEGRATED FRAMEWORK FOR SECURE IOT-BASED SMART VEHICLE MANAGEMENT USING MACHINE LEARNING AND BLOCKCHAIN FOR PATH PLANNING AND COLLISION AVOIDANCE
Author:	M V NARAYANA, P PARTHASARADHY, P JAGDISH KUMAR, SURYA KIRAN CHEBROLU, SANJEEV SHRIVASTAVA, S DILLI BABU, NILADRI SEKHAR DEY
Abstract:	The enormous developments taking place in technology for autonomous vehicles has created a big demand on intelligent systems retarding the zero accidents and fatalities transportation system needed for current highways. In this paper, we introduce a new integration framework incorporating IoT, machine learning and blockchain based technologies to overcome two major issues like path planning and collision avoidance in smart vehicle management. The framework will make use of IoT devices to collect data in real-time and communicate it between vehicles as well as with roadside infrastructure. The dynamic path planning for the execution of real-time decisions based on roadway terrain with different prediction and avoidance is optimized through machine learning algorithms. This immense quantity of exchanged data has been disrupted to be managed with a reliable standard in a decentralized, tamper-proof method using blockchain technology sunders the core trait of how IoT ecosystem operates. Using blockchain, the framework also guarantees vehicle-to-vehicle (V2V), vehicle to traffic light control system and then a vehicle computing server such as edge/fog/cloud servers so that all communications between vehicles by providing security and data sharing environment with confidence. The feature of blockchain integration does not only reduces the risks of hacking and tampering with data, but it also increases transparency as well as accountability in this system. The performance of the proposed framework is then examined through large-scale simulations and field experiments in different highway scenarios. Results show significantly improved path optimization, accident prevention and system wide reliability. The study shows that this holistic solution can help reduce traffic jams, improve road safety and support the growth of autonomous vehicles. This work adds to a larger collection of research efforts around smart transportation systems and underscores the utility in integrating novel methodologies towards establishing an improved vehicular environment.
Keywords:	Smart Vehicle Management, IoT, Machine Learning, Blockchain, Path Planning.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	A SYSTEMATIC REVIEW OF HEALTHCARE MONITORING SYSTEM: METHODS, APPLICATIONS, AND CHALLENGES
Author:	ADHAM M. SAEED, YASAMIN H. ALAGRASH, MERIAM JEME, L. BEN SAID
Abstract:	The rapid integration of the Internet of Things (IoT) in healthcare has paved the way for intelligent Healthcare Monitoring Systems (HMS) capable of real-time patient tracking, early diagnosis, and personalized treatment. This systematic review evaluates 180 studies, with 38 selected for in-depth analysis, focusing on technologies, applications, and challenges in HMS. The paper explores the architecture and operation of HMS, especially those empowered by Body Sensor Networks (BSNs), cloud computing, and machine learning. It identifies key application areas such as remote patient monitoring, wearable sensor integration, and secure data transmission, while also discussing gaps in interoperability, scalability, privacy, and energy efficiency. The findings emphasize the need for robust, privacy-respecting, and adaptive systems that support real-time performance and intelligent decision-making. Ultimately, this study contributes a comprehensive classification of current approaches, highlights technological trends, and proposes future research directions to guide the development of efficient and trustworthy healthcare monitoring systems.
Keywords:	Internet of Things (IoT), Healthcare Ubiquitous Healthcare (u-health), Real-Time Monitoring, Body Sensor Networks, Machine learning algorithms.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	PROFESSIONAL MOBILITY OF TEACHERS THROUGH IT-DRIVEN DIGITAL TRANSFORMATION OF EDUCATION
Author:	SOFIYA CHOVRIY, IRYNA ZHOROVA, OLHA KHUDENKO, SVITLANA STEBLIUK
Abstract:	The studys relevance is due to the growing role of digital technologies in education and the need to understand their impact on the professional mobility of teachers. The study aims to determine the effect of the use of digital platforms on teachers' professional mobility level in the context of participation in academic mobility programs. The research methodology included questionnaires, statistical analysis, correlation and factor analysis. Validity was ensured by using Cronbachs α and social desirability testing. The hypothesis was that digital platforms contribute to the growth of teachers' professional mobility and professional development. The results showed that teachers who use 4-5 digital platforms participate in 5.6-6.2 international events on average. Users of only one platform participated in 2.3. Participation in international projects among users of digital platforms is 48.2% (compared to 22.3% among supporters of traditional education), and participation in conferences is 55.4% (compared to 30.1%). The most influential platforms are Zoom (72.5%), Google Classroom (68.2%), and Microsoft Teams (62.1%). The authors conclude that digital platforms facilitate access to new teaching methods and significantly expand opportunities for international cooperation. The scientific novelty lies in a comprehensive analysis of the impact of digital platforms on various aspects of professional mobility using factor analysis. Prospects for further research are related to the study of barriers to digital integration and the development of recommendations for increasing the digital mobility of teachers in countries with different levels of digital infrastructure.
Keywords:	Digital Platforms, Professional Mobility, Integration Of Education, Distance Learning, Online Education, Academic Mobility, Digital Transformation Of Education, Digital Competence, Pedagogy, Knowledge, Sustainability Education, Education For Sustainable Development, Accessible Education, Tertiary Education.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	A ROBUST EXPLAINABLE RECURRENT DEEP Q LEARNING FOR DETECTING MULTICLASS INTRUSIONS IN IOT
Author:	PADMASRI TURAKA, SAROJ KUMAR PANIGRAHY
Abstract:	The deployment of explainable Artificial Intelligence (XAI) in cybersecurity applications is gaining much attention these days, but more research is still needed to better understand how effective it is at identifying attack exteriors and trajectories. The growing applications of machine learning (ML) and deep learning (DL) models in cyber defense, particularly anomaly-based Intrusion Detection Systems (IDS), necessitate the interpretation and the justification of their predictions to investigate potential cyberattacks. Hence, in this framework, an effective explainable DL technique is introduced to classify multiple intrusion attacks in the IoT-WSN system. Initially, the input data are taken from the UNSW-NB15 dataset, and then Weighted Correlated Adaptive Min-Max normalization (WCAdapt-MMN) is utilized to remove unwanted null values. Finally, the pre-processed data are fed as input into a Recurrent Deep Q learning (RDQN) model for intrusion classification as Fuzzers, DoS, Exploits, Generic, Reconnaissance, Shellcode, and Normal. Furthermore, four XAI models are investigated for enriched visualizations over the IDS: Shapley additive explanations (SHAP), Permutation Feature Importance, and Partial Dependence Plot (PDP). A Python simulation tool is utilized for the simulation process, and a freely accessible UNSW-NB15 dataset is considered for the training process. In the resultant section, the accuracy of 98.48%, Matthew’s correlation coefficient (MCC) of 0.98, precision of 98.62%, recall of 98.6%, False positive rate (FPR) of 0.319, and computation time (CT) of 49s for enhancing the robustness against IoT attacks.
Keywords:	Intrusion Attack; Multi-Attack Detection; Internet of Things; Explainable Artificial Intelligence; Recurrent Deep Q Learning; Shapley Additive Explanations; Permutation Feature Importance; Partial Dependence Plot
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	PRIVACY-PRESERVING CLOUD DATA MINING USING HOMOMORPHIC ENCRYPTION-BASED LOGISTIC REGRESSION
Author:	ANKITA SINGH , ANNA ALPHY
Abstract:	Because scalable, secure machine learning in untrusted cloud environments is critical today, existing privacy-preserving approaches—such as anonymization and differential privacy—often compromise model utility, and prior homomorphic encryption (HE) solutions for logistic regression suffer from high computational overhead, partial encryption, or poor scalability. To address these limitations, we propose a fully homomorphic encryption (FHE)-based logistic regression framework that ensures end-to-end data confidentiality without decryption and supports fully encrypted training and inference. We evaluate our approach on the UCI Heart Disease dataset, achieving 84.2% classification accuracy, with less than 2% degradation compared to plaintext models, while substantially reducing computational overhead relative to recent HE-based techniques. Our analysis explores privacy-accuracy-computation trade-offs and demonstrates a practical and efficient balance between these competing objectives. These results indicate that FHE-enabled logistic regression can deliver high accuracy in privacy-sensitive domains such as healthcare without prohibitive performance costs. This framework paves the way for broader adoption of secure, privacy-preserving machine learning on cloud platforms, offering a dependable solution for organizations seeking to analyse sensitive data securely.
Keywords:	Privacy-preserving machine learning, Fully homomorphic encryption (FHE), Logistic regression, Cloud data mining, Secure computation, Encrypted training, Data confidentiality, Privacy-awareness
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	DIMENSIONALITY REDUCTION BASED ON ENSEMBLE CLUSTERING APPROACH
Author:	SWETHA T, SAKTHIVEL G, SANDHYA RANI D
Abstract:	The dimensionality reduction is still a difficult problem due to the vast volumes of data generated in social networks and bio- informatics. Here, a new dimensionality reduction approach called CCBFS-FR is devised with the inspiration of CCBFS method. This algorithm leverages genetic algorithm and is based on ensemble clustering or ensemble clustering concept. Initial population of genetic algorithm is generated using five feature rankers. A popular dissimilarity measure called Symmetric Distance Difference (SDD) is used as fitness function as similar to CCBFS. As genetic algorithm uses some stopping criteria to limit the number of populations, we use stopping criteria similar to CCBFS which was proved to be good one. The time complexity of CCBFS-FR is also less compared to CCBFS. The technique is tested on most challenging benchmark datasets with dimensionality ranging from low to medium using classification as validation. Except for Breast cancer data set we achieve good accuracies and less number of features for remaining data sets compared to literature. For Wine data set we achieve the accuracy as 96.28% with 7 features, PIMA achieved 80.35% with 4 features and Wine quality white achieved 68.72 with 7 features.
Keywords:	Dimensionality Reduction, Ensemble Clustering, Genetic Algorithms
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	ADAPTIVE MULTI-PATH FUSION: A NOVEL MULTIMODAL LEARNING FRAMEWORK WITH UNCERTAINTY-AWARE ROUTING
Author:	DR. HARI JYOTHULA, DR. SUBBA RAO POLAMURI, DR. G SIVA KRISHNA, DR SYAMALA RAO,MV RAJESH, B SATYANARAYANA MURTHY
Abstract:	Multimodal learning is increasingly vital in real-world AI tasks where information originates from diverse and often unreliable sources. However, most existing methods employ rigid fusion strategies that fail under noisy, missing, or irrelevant modalities. This study introduces Adaptive Multi-Path Fusion (AMP-Fusion), a novel multimodal learning framework with uncertainty-aware routing that dynamically selects the most reliable information paths. Our hypothesis is that adaptively weighting modalities based on informativeness and uncertainty will significantly improve robustness compared to static fusion methods. Large-scale experiments in vision-language, audio-visual, and healthcare tasks confirm our hypothesis, showing AMP-Fusion achieves up to 15% improvement over state-of-the-art models while maintaining over 91% performance even when 40% of modalities are missing or noisy. These findings highlight AMP-Fusion’s importance for practical, real-world multimodal AI applications where data reliability cannot be guaranteed.
Keywords:	Multimodal Learning, Adaptive Fusion, Attention Routing, Uncertainty Quantification, Missing Modalities, Robust AI, Transformer-based Models.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	AIOT-DRIVEN WEARABLE ARCHITECTURE FOR INTELLIGENT GRAPHIC DESIGN AUTOMATION
Author:	TANAPEAK PEXYEAN, PANIT THONGDEE, CHUTIMA KETSA, NATTAPON LONIN, PRACHYANUN NILSOOK, MAYKIN WARASART
Abstract:	The purpose of this research is to develop a graphic design system through automated wearables and to evaluate the suitability of a graphic design system through automated wearable devices. Phase 2 is the development of a graphic design system through automated wearable devices, which consists of Graphic Design Automation that works with smart wearable devices to support and optimize the graphic design process. The system consists of four main layers: the data import layer, the data processing layer, the integration layer, and the application layer, where the data from the user is transmitted through the wearable device to the processing layer, which processes data at both the edge computing and cloud computing levels using machine learning (ML) and artificial intelligence (AI) technologies such as gesture recognition and natural language processing (NLP). The system also provides user feedback through haptic feedback and user data analysis to improve the accuracy of automated commands in the future. In addition, data encryption is provided for security during data transmission as a new approach to digital design development by integrating wearable technology with an Intelligent Automation System. It improves ease of use and simplifies technical complexity for designers. The system can also be connected to professional graphic design software to increase productivity and create automated designs based on real-time sensor data. AI-driven design integrates IoT and user-centric design automation to increase speed, flexibility, and responsiveness to the needs of digital content creation. This integration facilitates efficiency, flexibility, and convenience in the design process. Supports the development and integration of various software effectively.
Keywords:	Wearable Devices, Artificial Intelligence Internet of Thing, Graphic Design Automation
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	WORD VECTORS FOR CRITICISM OF A KOREAN FILM ‘DECISION TO LEAVE’ BY CHANWOOK PARK
Author:	KWANGHO KO , JURON PAIK
Abstract:	This research presents a novel approach to film analysis, leveraging word vectors to objectively evaluate movie critiques. Focusing on Director Chanwook Park's award-winning film, 'Decision to Leave', the study employs word vectors derived from the movie's script of Korean text. Traditional critiques often emphasize contrasting elements and themes, but their subjective nature poses challenges in objective validation. To address this, we trained a language model using LSTM on the film's script, obtaining word vectors that capture the essence of the narrative. These vectors were then used to perform various text analyses, including similarity and analogy operation for the keywords suggested by the critiques. By comparing the semantic relationships in the critiques with those derived from the word vectors, we could objectively validate the critiques' assertions. Furthermore, we visualized the word vectors in a two-dimensional space, confirming the spatial relationships of key terms highlighted in critiques. The study underscores the potential of word vectors in providing a more objective lens for film analysis, bridging the gap between traditional film criticism and data-driven insights.
Keywords:	Analogy Preparation, Chanwook Park, Film Criticism, Language Model, Word Vector
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	OPTIMIZED SERVERLESS FRAMEWORK FOR SCALABLE DEPLOYMENT OF DEEPAR+ FORECASTING
Author:	GANESHAN MAHALINGAM, RAJESH APPUSAMY
Abstract:	This study proposes an optimized serverless framework for the scalable and cost-efficient deployment of DeepAR+ forecasting models, addressing critical challenges in time-series inference such as cost overhead, latency and operational complexity. The architecture integrates AWS SageMaker Batch Transform and OpenWhisk containerized actions to enable on-demand, low-cost model execution, achieving up to 40% reduction in operational expenses compared to always-on endpoints. To minimize latency, precomputed DeepAR+ predictions are cached in Redis, and Apache Kafka decouples heavy inference workloads from real-time processing, allowing sub-second anomaly detection. Security is strengthened through VPC-isolated deployments, IAM-based access control and real-time auditing via CloudTrail, improving unauthorized access detection to 99.9%. A lightweight pre-processing pipeline ensures consistent input data, even in sparse telemetry environments. The proposed solution delivers a secure, scalable and efficient forecasting framework that leverages the flexibility of serverless infrastructure while maintaining real-time responsiveness, making it well-suited for modern predictive analytics applications in data-intensive environments.
Keywords:	DeepAR+, Serverless, SageMaker, AWS IoT Core, AWS Glue, OpenWhisk
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	A PROPOSED FEDERATED LEARNING (FL) APPROACH TO ENHANCE PREDICTION OF STUDENT PERFORMANCE EVALUATION SYSTEMS
Author:	TAHER ABOZAID ELSNOUSY, MOHAMED RAGAIE SAYED, NASHAAT ELKHAMISY ELGHITANY
Abstract:	It is essential to forecast student performance accurately in order to support early intervention and tailored learning. Centralized data gathering is the foundation of traditional machine learning (ML) techniques, which restricts inter-institution collaboration and increases privacy issues .By facilitating decentralized model training without sharing raw data, Federated Learning (FL) offers a privacy-preserving substitute. The centralized ML and FL models used to predict student performance are compared in this study. An educational benchmark dataset was pre-processed, divided into several clients, and assessed in both non-IID and independent and identically distributed (IID) scenarios. Using Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and R2 score, models such as Linear Regression, Elastic Net, Random Forest, Gradient Boosting, Support Vector Regression (SVR), and Multi-Layer Perceptron (MLP) were trained and evaluated. The findings show that FL can maintain data privacy while achieving competitive performance. The exceptionally substantial performance gap that was observed necessitates careful interpretation and emphasizes the significance of meticulous pre-processing, baseline fairness, and statistical validation, even when FL models exhibited lower error metrics than centralized ML. In line with theoretical predictions, experiments also demonstrated that FL performance is negatively impacted by non-IID client distributions. A comprehensive dataset and partitioning description, a transparent methodology including statistical testing, and a reproducible FL assessment framework are all contributions of the study. Results highlight FL's potential in educational settings, but they also highlight its drawbacks and the necessity of more research on scalability, robust privacy safeguards, and model customization.
Keywords:	Federated Learning, Machine Learning, Student Performance Prediction,Educational Data Privacy.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	DIGITALIZATION OF THE FINANCIAL SYSTEM IN THE CONTEXT OF ENSURING NATIONAL SECURITY
Author:	SERHIY SHKARLET, MAKSYM DUBYNA, ROMAN REUS, DENIS RINZHUK, VLADYSLAV ZELENSKYI, ARTEM MALYKHIN
Abstract:	In the article, the role of digitalization of the financial system in ensuring the national security of the country is considered. This is implemented through the study of the national security structure as an integral system, isolation of the role of financial security in formation of stable development of the country, description of its significance for formation of the national security of the country. This, in turn, it is possible to state that financial security is a characteristic of the specific state of the financial system functioning of the country and it is substantiated in the article that there is relationship between this system and the financial security level. In the future, the structure of the financial system was considered as a system, its main components were determined, and formation of prerequisites was specified. To consider digitalization features of financial systems in the article, the essence of digital transformations was considered, prerequisites of these transformations of financial systems were defined. In the future, it is specified that first digitalization of financial systems happens through digital transformation systems of public finance and active using of digital technologies within the domestic financial market. In the article, the role of the public finances system in providing national economy was determined in detail, positive and negative consequences of digital transformations of these systems were defined. Characteristic traits of the financial market functioning within the national economy were also identified, features of its digital transformations were defined, namely positive and negative consequences its digitalization were highlighted. As a result, constructive and destructive consequences of digitalization of financial systems to ensure national security were determined.
Keywords:	Digitalization, Financial System, National Security, Public Safety Finance System, Financial Market, Financial Institution, Banking Institution, Financial Safety, Economic Security.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	REVEALING BIAS DIFFUSION IN SOCIAL MEDIA THROUGH STRUCTURAL AND SEMANTIC MEASURES
Author:	ELIZABETH LEAH GEORGE , SUBASHINI PARTHASARATHY
Abstract:	Social media has become a powerful force shaping public narratives, particularly on sensitive topics. However, these platforms also allow biased or unfair language to spread quickly. This study examines the diffusion of biased language on violent crime through Twitter discourse (using the hashtag #Murder). Understanding such bias diffusion is crucial because it can skew public perception and potentially impact judicial outcomes. Our goal is to measure and mitigate bias by combining advanced language models and network analysis. We propose a framework that integrates semantic language processing and structural topology analysis. A transformer-based model (T5) extracts semantic features from tweets, and to test bias mitigation, we apply text style transfer to rewrite content into neutral, positive, or negative forms. We build a two-layer multiplex concept network: one layer for general (unbiased) concepts from term co-occurrence and another for bias-related concepts identified via a curated lexicon of sensitive terms links between layers capture where neutral and bias-laden concepts co-occur, revealing bias infiltration pathways. We quantify structural and semantic similarity using Jaccard similarity, graph edit distance, cosine similarity, DeltaCon, and NetSimile metrics, and we measure bias propagation using a bias infiltration index. Bias-related terms are fully integrated into the concept network in the original tweet corpus (Jaccard node overlap = 1.0, Weighted Jaccard = 1.0), indicating that biased language co-occurs with all key topics. After tweets are rewritten by sentiment, this overlap decreases: the overall node overlap drops to ~0.90, and the weighted overlap declines most in neutral rewrites (to ~0.76, versus ~0.88 in positive/negative cases). Positive tweets embed bias most strongly, whereas neutral rewrites retain a high degree of subtle bias (as indicated by the Bias Infiltration Index). Statistical tests (paired t-tests, p > 0.49) confirm that the reductions in explicit bias are not significant, underscoring the bias’s resilience. Overall, our findings suggest that while sentiment-based rewriting can partly disrupt bias pathways, biased concepts remain deeply woven into social media discourse.
Keywords:	Social Media Discourse, Bias Detection, Semantic Embeddings, Text Style Transfer, Concept Maps
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	ASSESSING THE EFFECTIVENESS OF CHATGPT-MEDIATED INSTRUCTION IN DEVELOPING PARAGRAPH WRITING SKILLS
Author:	TAJ MOHAMMAD, ALI ABBAS FALAH ALZUBI, MOHD NAZIM, SOADA IDRIS KHAN
Abstract:	The integration of Artificial Intelligence (AI) tools into education has recently captured the attention of educators and researchers, as they have the potential to improve teaching and learning. Among these tools, ChatGPT has become a key asset for writing instruction. However, its impact on discrete writing skills such as paragraph development is not well studied. This study investigates the effectiveness of ChatGPT-mediated instruction in enhancing the paragraph writing skills of English as a Foreign Language (EFL) students. Using a quasi-experimental design, the research included two groups: an experimental group (N = 30) that received writing instruction with ChatGPT support, and a control group (N = 30) that received conventional instruction. Data were gathered using pre- and post-tests to assess growth in paragraph writing skills and semi-structured interviews to document students’ experiences with ChatGPT-supported instruction. The results indicated that the experimental group achieved notable gains in paragraph writing. The findings suggest practical steps for EFL teachers to integrate ChatGPT as an innovative tool to strengthen students' writing competence. Additionally, curriculum designers should create clear guidelines and professional development to facilitate these applications.
Keywords:	ChatGPT-Mediated Instruction, Paragraph Writing Skills, Experimental Research Design
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	NEURAL REPRESENTATIONS FOR 3D RECONSTRUCTION (2017–2025): TAXONOMY, BENCHMARKS, AND HYBRID TRENDS
Author:	FENG WENWEN, SITI KHADIJAH ALI, CEN XIAO, RAHMITA WIRZA O.K. RAHMAT
Abstract:	Three-dimensional (3D) reconstruction remains a core challenge in computer vision due to the inherent ambiguity of recovering structure from two-dimensional (2D) imagery. Traditional geometry-driven pipelines, such as multi-view stereo and structure-from-motion, achieved notable success but relied on handcrafted features and strict imaging conditions, limiting robustness and scalability. Over the past decade, deep learning has transformed the field by enabling direct inference of geometric priors. However, persistent trade-offs between fidelity, efficiency, and generalization highlight the need for a systematic synthesis of advances. This survey reviews deep learning approaches for 3D reconstruction between 2017 and 2025, structured around five major shape representation paradigms: voxels, point clouds, meshes, implicit surfaces, and wireframes. For each, we analyze representative methods, benchmark datasets, and evaluation metrics, emphasizing their respective strengths and limitations. Beyond single paradigms, we examine hybrid and cross-representation frameworks that integrate convolutional neural networks, transformers, and generative models to exploit complementary advantages. The key contributions of this work are threefold: (i) establishing a taxonomy across five paradigms with temporal coverage from 2017–2025, (ii) synthesizing emerging hybrid and transformer-driven frameworks as unifying approaches, and (iii) identifying open challenges in scalability, domain generalization, and perceptual quality. By bridging earlier geometry-centric surveys with recent transformer-based and pretraining-driven advances, this survey generates new knowledge on the evolution and convergence of neural representations. It provides researchers and practitioners with a roadmap toward scalable, accurate, and semantically enriched 3D reconstruction systems capable of addressing real-world complexity.
Keywords:	3D Reconstruction, Deep Learning, Shape Representation, Hybrid Representation, Multi-Modal Fusion
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	AI-POWERED DIAGNOSIS OF NEONATAL RESPIRATORY DISTRESS SYNDROME: A COMPARATIVE STUDY OF CNN MODELS ON CHEST X-RAYS USING CLINICAL MOROCCAN DATASET
Author:	SALMA SEKKAT, MOUNA ZOUINE, ZINEB CHEKER, OUSSAMA FANGACHI, KARIMA SAMMOUD, SAAD CHAKKOR, ABDALLAH OULMAATI, ADIL NAJDI
Abstract:	Early and accurate diagnosis of Neonatal Respiratory Distress Syndrome (RDS) is vital for improving outcomes in preterm infants. Traditional diagnostic methods often suffer from subjectivity and limited scalability, especially in resource-constrained settings. This study investigates the potential of convolutional neural networks (CNNs) for automated RDS detection from chest X-rays images, using a clinically annotated dataset sourced from multiple Moroccan hospitals. A comparative analysis of seven CNN architectures (InceptionV3, ResNet-50, DenseNet121, VGG16, EfficientNetV2, ConvNeXt, and a custom lightweight CNN) was conducted. Models were evaluated using precision, recall, F1-score, accuracy, and AUC. InceptionV3 achieved the highest performance with 93% accuracy and 96% recall for RDS, while the custom CNN demonstrated strong sensitivity (95% recall), making it well-suited for deployment in low-resource environments. The study underscores the effectiveness of deep learning for neonatal diagnostics and highlights the importance of context-specific AI solutions in enhancing early detection and decision support for RDS in clinical practice.
Keywords:	RDS, Chest X-rays Classification, Neonatology, Diagnostic Support System, CNNs, Deep Learning, InceptionV3, EfficientNetV2, ConvNeXt.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	AN ENHANCED EDGE PRESERVING TECHNIQUE FOR MULTIFOCUS IMAGE FUSION
Author:	VAKAIMALAR ELAMARAN, RAMYA R, ANANDH A, PRAVEEN KUMAR PREM KAMAL, DEEPA PRIYA V
Abstract:	Image fusion aims to combine complementary and repetitive information present in multiple images of the same scene that has been captured by different cameras or sensors and generates a single image with improved quality. This work proposes an algorithm for fusing multifocus images using edge superimposition and Dense Scale Invariant Feature Transform (Dense SIFT). Due to the depth of the field, when background is focused, boundary between the background and expanded foreground objects become unclear and vice versa. To overcome this, edge superimposition is applied to enhance the boundary region between the foreground and background objects. As an initial step, edges are detected from a source image using sobel operator and the extracted edge features are superimposed on the next source image and vice versa. This step plays a vital role in the proposed approach, since sharpness of the edges is crucial in the picturesque quality of the image. After super imposition of edges, Dense SIFT descriptor extracts features at every pixel and helps in deciding the resultant pixels in the fused image. The image feature-based metrics such as Average Gradient, Edge Intensity, Standard deviation and Spatial Frequency demonstrates the superior ability of the proposed algorithm in preserving edges compared to numerous other algorithms under consideration.
Keywords:	Dense SIFT, Feature, Image fusion, Multifocus, Superimposition.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	PREDICTION OF CRASHES IN MOEX RUSSIAN INDEX USING LOG-PERIODICITY
Author:	AMMAR HAVELIWALA, SURYANSH SUNIL, PALANIAPPAN SELLAPPAN, RAJESH MAHADEVA, VARUN SARDA
Abstract:	Financial crises in emerging nations pose significant dangers to economic stability yet predicting them remains difficult. This study tests whether speculative bubbles and subsequent crashes can be predicted by using the Log-Periodic Power Law (LPPL) model to the MOEX Russian Index using daily data from 1997–2024. We estimate that with little lag, LPPL can predict crash dates and detect important moments of instability in MOEX. The results demonstrate that log-transformed data produce smaller error margins, indicating trade-offs between variance capture and forecasting precision, even if raw price data frequently anticipate crashes more accurately in terms of timing. Although LPPL's ability to predict smaller collapses and lower false positives is still limited, these findings imply that it can give regulators and policymakers early warning signs of systemic risk.
Keywords:	LPPL model, Financial Market, Speculative Bubbles, MOEX Russian Index, Market Crashes
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	ENHANCING WEED CLASSIFICATION IN AGRICULTURE: LEVERAGING DEEP NEURAL NETWORKS AND SEGMENT ANYTHING MODEL
Author:	PRASANNA DRL, DIVYA LINGINENI, MANISAIGANESH KOTHA, QAWIUDDIN MOHAMMED
Abstract:	The paper entails the utilization of deep neural network models such as ResNet50, and Inception V3 with a state-of-the-art segmentation model Segment Anything Model from Facebook for enhancing plant identification. The primary objective is to enhance weed identification within agriculture through the utilization of Deep Neural Networks as well as the Segment Anything Model. The study discusses how enhanced weed species identification accuracy as well as precision through advanced segmentation techniques can be achieved. The study used Segment Anything Masks (SAM) for image augmentation, preprocessing techniques of neural networks, as well as the structure of the neural network for classifying. It further used comparative modeling of models such as DenseNet169, ResNet-50, Inception-v3, as well as VGG19 for optimizing weed control strategies. The project design as well as evaluation outcome is of a deep learning-enabled weed classifier as well as advanced image augmentation suggested techniques of well-defined plant bounds, comparison with other alternatives for optimal weed control. The outcome entails higher accuracy of weed species identification as well as the possibility of precision agriculture. Through the combination of powerful segmentation technologies with powerful deep-learning technologies, the project efficiently enhanced the accuracy of identification as well as classifying of the weed plants.
Keywords:	Deep Neural Networks, Segment Anything Model (SAM), Weed Identification, Image Augmentation, Precision Agriculture.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	LEAFHEALTHAI: AN ROI-OPTIMIZED DEEP LEARNING FRAMEWORK FOR ENHANCED LEAF DISEASE DETECTION IN PRECISION AGRICULTURE
Author:	B. SUREKHA, DR. T. SUBHAMASTHAN RAO
Abstract:	Early and accurate detection of plant leaf diseases is essential for improving crop yield and ensuring sustainable agricultural practices. Traditional Convolutional Neural Network (CNN) models often process entire leaf images without emphasizing disease-affected regions, leading to misclassifications, poor interpretability, and reduced performance under real-world conditions. This study introduces LeafHealthAI, a novel deep learning framework integrating Grad-CAM-based Region of Interest (ROI) attention with a weighted categorical cross-entropy loss to enhance precision in disease localization and classification. The proposed LDDNet-WROI model leverages ROI-guided learning to focus on disease-relevant features while dynamically addressing class imbalance. Experiments conducted on the benchmark PlantVillage dataset, comprising 54,000 images across 38 disease classes, demonstrate that LeafHealthAI achieves a classification accuracy of 98.46%, surpassing baseline models such as VGG16, ResNet50, and EfficientNetB0. Visualization using Grad-CAM heatmaps confirms the framework’s interpretability by clearly highlighting disease-affected areas. The results validate that ROI-driven attention and weighted optimization significantly improve model robustness, making LeafHealthAI suitable for real-time precision agriculture applications and scalable deployment in IoT-enabled crop monitoring systems.
Keywords:	Plant Disease Detection, Deep Learning, Region of Interest, Weighted Loss, Precision Agriculture
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	DETECTING MENTAL ILLNESS THROUGH SENTIMENT ANALYSIS ON SOCIAL MEDIA USING ENSEMBLE MACHINE LEARNING
Author:	S V N SREENIVASU, MR.V.S.N. MURTHY, E. VIJAYA LAKSHMI, J L SARWANI THEEPARTHI, G SATYANARAYANA, B K MADHAVI, G. SIVAKUMAR
Abstract:	The Research proposed in this paper concentrate on the discovery and classification of mental disorders are using sentiment analysis techniques over socialmedia data. Platforms like Reddit and Twitter are rich in user-generated content, many of which are related to mental and psychological well-being. The research utilized the CLPsych 2015 Shared Task Dataset, which comprises social media posts tagged with various mental health conditions (depression, PTSD, anxiety, and bipolar disorder). By combining linguistic markers through Natural Language Processing (NLP) with sentiment evaluation components, the study seeks to identify the emotional nuances and behavioral cues linked to these disorders. In addition, the efficacy of various ensemble machine learning models such as Random Forest, Gradient Boosting, and XGBoost were compared to classical classifiers: Logistic Regression, Naive Bayes and Support Vector Machines (SVM) for multi-class classification. Measurement was carried out using an accuracy, precision, recall, and F1-score metrics. The ensemble methods all consistently outperformed individual methods, especially in addressing overlapping symptoms across mental health categories. This study demonstrates the potential of combining sentiment analysis with powerful ensemble learning strategies in early, automated and scalable identification of mental illnesses from social media contents.
Keywords:	Mental Illness Detection; Sentiment Analysis; Social Media; Multi-class Classification; Ensemble Learning; Natural Language Processing (NLP); Reddit; CLPsych 2015 Dataset; Depression; Anxiety; PTSD; Bipolar Disorder; Machine Learning; Text Mining.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	A FEDERATED LEARNING BASED APPROACH FOR CARDIOVASCULAR DISEASE PREDICTION
Author:	BISNA N D , AJAY JAMES ,HELNA THOMAS
Abstract:	Cardiovascular diseases refer to a group of conditions that affect the heart and blood vessels. Cardiac arrhythmia, specifically Atrial Fibrillation, is the most common and sustained type, and is associated with substantial morbidity. Detecting atrial fibrillation at an early stage is crucial as it is an early sign of most of the critical heart diseases such as stroke, heart failure, and other life-threatening cardiovascular complications. Accurate AF detection enables timely interventions and significantly improves the prevention of critical heart diseases, reducing morbidity and mortality. The increasing prevalence of chronic heart diseases requires advanced predictive techniques for early intervention and personalized healthcare, but the sensitivity of health data raises privacy concerns. This study proposes a privacy-preserving deep learning framework for multiclass heart disease prediction using electrocardiogram signals. To preserve data privacy and ensure decentralized model training, federated learning is employed with multiple strategies including Federated Averaging, Federated Stochastic Gradient Descent, and random client participation. The Deep Neural Network model is optimized using the Adam optimizer and trained with sparse categorical crossentropy loss to handle multi-class classification effectively. Experiments showed that FedAvg improves with more clients and rounds, while FedSGD maintains stable accuracy and outperforms FedAvg with higher client counts. Optimization techniques such as quantization reduce memory usage, and knowledge distillation improves the performance of the compressed model thus making it suitable to deploy in rsource constrained environment. This study aimed to efficiently predict chronic heart diseases while safeguarding sensitive health information using federated learning.
Keywords:	Cardiovascular Diseases, Atrial Fibrillation, Machine Learning, Deep Learning, ElectrocardioGram, Federated Learning, Quantization, Knowledge Distillation, FedAVG, FedSGD
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	AUTOMATED DETECTION OF MULTIPLE SCLEROSIS LESIONS IN MRI USING AN EXPLAINABLE AI FRAMEWORK WITH HIERARCHICAL FEATURE FUSION
Author:	DR.A.AZHAGAMMAL, T.KARTHIKEYAN, KROVVIDI S B AMBIKA, JAKKAPU NAGALAKSHMIDEVI, RAYAVARAPU SRIDIVYA, DR.T.VENGATESH, DR.S.S.ANANTHAN, POTHUMARTHI SRIDEVI
Abstract:	This paper introduces XAI-FuseNet, a novel explainable AI framework designed to address the challenges of multi-modal data fusion and interpretability in the automated segmentation of Multiple Sclerosis (MS) lesions from MRI scans. The proposed model is based on an encoder-decoder architecture integrated with two key components: a Hierarchical Feature Fusion (HFF) module that dynamically combines multi-scale features from different MRI sequences (e.g., T1-w, T2-w, FLAIR), and an integrated attention mechanism that guides the segmentation process while generating high-resolution saliency maps for visual explainability. Evaluated on the MSSEG-2016 dataset, XAI-FuseNet achieved a Dice similarity coefficient of 78.5% and a lesion-wise F1-score of 81.2%, outperforming strong baselines including U-Net (72.1%) and nnU-Net (76.8%). Furthermore, the visual explanations produced by the model were validated by expert neuroradiologists, achieving a 92% concordance rate with clinical ground truth. These results demonstrate that XAI-FuseNet not only enhances segmentation accuracy through effective multi-modal fusion but also provides clinically interpretable explanations, thereby offering a trustworthy tool for AI-assisted diagnosis of MS.
Keywords:	Explainable AI (XAI), Multiple Sclerosis, MRI, Deep Learning, Semantic Segmentation, Feature Fusion, Attention Mechanisms, Medical Image Analysis.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	INTEGRATION OF DIGITAL PLATFORMS AND IT INFRASTRUCTURE TO INCREASE THE EFFECTIVENESS OF MARKETING STRATEGIES
Author:	HASSAN ALI AL-ABABNEH, TARIQ ABDELHAMID ALI MUSSALAM
Abstract:	In the context of the rapid development of digital technologies, the integration of digital platforms and IT infrastructure is becoming a key factor in increasing the effectiveness of marketing strategies of modern companies. The purpose of this study is to identify the impact of the integration of digital solutions on the optimization of marketing processes and strengthening the competitive advantages of a business. The relevance of the topic is due to the growing demand for complex IT tools that can provide flexibility, personalization and efficiency of marketing campaigns in a dynamic market. The study emphasizes the need for an integrated approach to synchronizing various digital platforms - from CRM and analytical systems to customer interaction channels - with the IT infrastructure of an organization. The research methodology is based on a combination of qualitative analysis of cases of leading companies and quantitative analysis of data on the effectiveness of marketing activities, including conversion, engagement and return on investment. The results demonstrate that the integration of digital platforms with IT infrastructure helps to increase targeting accuracy, reduce operating costs and speed up decision-making in marketing. The implementation of unified data management systems and process automation allows companies to quickly adapt marketing strategies to changes in the market and consumer preferences. The findings of the study confirm that successful integration of digital solutions is an integral part of modern marketing and significantly improves the quality of interaction with the target audience. The results can be used to develop recommendations for improving IT infrastructure and increasing the effectiveness of marketing strategies in various industries.
Keywords:	Digital Platforms, IT, Marketing, Integration
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	FEDERATED LEARNING-BASED DECENTRALIZED ENCRYPTION KEY GENERATION WITH ENHANCED CLOUD PRIVACY PROTECTIONS FOR MEDICAL DATA
Author:	HIMANSHU , PUSHPENDRA SINGH
Abstract:	The increasing digitization of healthcare data has underscored the urgent need for privacy-preserving, scalable, and efficient machine learning frameworks. This paper presents a novel architecture that combines Federated Learning (FL) with Differential Privacy (DP) and Decentralized Encryption Key Generation (DEKG) to protect sensitive patient data during collaborative model training. Unlike traditional centralized approaches that risk data leakage and violate compliance regulations, our method ensures that raw data never leaves the source, maintaining both patient confidentiality and model utility. We evaluated our approach using a processed version of the MIMIC-III dataset, containing structured patient records across five categories. The FL-only model achieved an accuracy of ~94.8%, while the integrated FL+DP framework delivered a higher accuracy of 96.1% under a strict privacy budget of ε = 1.9, demonstrating a minimal trade-off between privacy and performance. The model converged effectively within 15 communication rounds, and client participation rates exceeded 85%, highlighting the system’s scalability and robustness in heterogeneous environments. Our evaluation includes comprehensive metrics such as communication efficiency, per-client accuracy, confusion matrix analysis, and noise sensitivity. Additionally, a radar chart visualization validates performance consistency across varying FL and DP parameters. The inclusion of decentralized key management further mitigates single points of failure and enhances cryptographic security.
Keywords:	Federated Learning (FL), Differential Privacy (DP), Decentralized Key Generation, Healthcare Data Security, MIMIC-III Dataset, Privacy-Preserving AI
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	ARTIFICIAL INTELLIGENCE INTEGRATION IN KAIZEN-ORIENTED CONTINUOUS IMPROVEMENT FOR SMART MANUFACTURING
Author:	CHAIMAE ADADI, HAKIM JEBARI, NAOUFAL SEFIANI
Abstract:	In an era marked by rapid technological advancement and pervasive connectivity, artificial intelligence (AI) is fundamentally transforming the digitalization of industrial processes. This transformation builds upon the foundations of Lean Manufacturing—a management philosophy centered on continuous improvement of processes, products, and services. Within the Industry 4.0 framework, the integration of AI significantly enhances the effectiveness of Lean tools by automating repetitive tasks, enabling real-time processing of vast data streams, and supporting faster, data-driven decision-making. By leveraging machine learning and predictive analytics, AI empowers industries to anticipate disruptions, optimize production workflows, and dramatically boost overall productivity. Moreover, embedding AI into continuous improvement (CI) cycles introduces powerful new dynamics. When combined with Kaizen principles, smart technologies accelerate innovation, streamline improvement timelines, and offer more agile responses to the growing complexity and demands of modern industrial ecosystems.
Keywords:	Artificial Intelligence, Continuous Improvement, Kaizen, Smart Manufacturing, Digital Transformation
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	COGNITIVE AI FOR TIME-AWARE MODELING OF BRAIN SIGNALS IN HEALTH AND PHYSIOLOGICAL SYSTEMS
Author:	B. RAMESH, DR. N. SATHEESH KUMAR, DR. E. ARAVIND
Abstract:	There is a need of advanced tools which are capable of identifying or interpreting the multiscale, nonlinear and complex patterns. The proposed cognitive AI offers an innovative and novel approach to address these kinds of challenges, inspired by human brain, learning and perception. In this paper we introduced cognitive AI based framework to analyze the brain dynamics with different time scales in applications like diagnosis, physiological modeling including health monitoring. The main objective is to study various datasets for experiments and to identify current developments in this field so that, it will show the path to design and implement cognitive models to integrate brain signals to represent scalable framework. The proposed approach integrates temporal deep learning structures such as LSTM, GRU and cognitive structure. The evaluation on real world datasets gives high accuracy and relevance for clinical use cases. Finally, this paper contributes to the evolution of intelligent health technologies with cognitive AI model for physiology interaction over the time period. The results shows that the accuracy of cognitive AI is 89% when we deal with large datasets, whereas the accuracy levels of LSTM and GRU found as 72% and 70% respectively. The other required parameters are demonstrated in the results section. The proposed framework for Cognitive AI has outperformed the other models in terms detecting or interpreting brain dynamics effectively across time scales to support health care applications.
Keywords:	Cognitive AI, Brain Dynamics, GRU, LSTM, Temporal Deep Learning
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	BAND SELECTION USING HYBRID OPTIMIZATION ALGORITHM FOR HYPERSPECTRAL IMAGE CLASSIFICATION
Author:	VINAY YADAV ESARAPU, PHANEENDRA KUMAR B L N
Abstract:	Band selection is an important processing method for hyperspectral image analysis due to data’s high redundancy and high dimensionality. The PSO-GWO hybrid utilizes the strengths of Particle Swarm Optimization (PSO) and the Grey Wolf Optimizer (GWO) to provide a trade-off between exploration and exploitation for hyperspectral band selection. PSO is effective in exploiting the search space by using personal and global best positions to update solutions, resulting in fast convergence towards good areas. Alternatively, GWO encourages exploration through grey wolf social behaviour imitation and hunting to avoid premature convergence and exploration of alternative solutions. Merging the two into a single paradigm, hybrid model utilizes PSO's convergence rate while adding the diversity control mechanisms of GWO to achieve more stable and accurate band selection in hyperspectral image classification. The selected bands are then trained using 3D CNN. The performance of the proposed method is tested via Indian Pines. The Overall Accuracy achieved with the framework is 98.35% whereas other methods it ranges between 95.24% and 95.72%. The present work can be used in Agriculture, Military and Defense applications.
Keywords:	Band Selection, Nature Inspired, Hyperspectral, Classification, Convolution Neural Networks
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	LANGUAGE DYNAMICS IN THE DIGITAL ENVIRONMENT THROUGH THE PRISM OF NEW MEANINGS AND CHANGED RULES
Author:	OLHA DEKALO, TETIANA KUZMENKO, MARYNA SHARAPA
Abstract:	In todays digital environment, communication is undergoing rapid changes that directly affect linguistic norms, communication styles, and the cultural function of language. The growing role of visual symbols, abbreviations, slang and anglicisms necessitates a scientific analysis of new linguistic realities. The purpose of the study is to identify the peculiarities of language transformations in digital communication on the example of five popular platforms. The methodological basis was a qualitative content analysis of messages in social networks, messengers, and forums. The study found that platforms focused on visual content (Instagram, Twitter) demonstrate the highest concentration of language changes, including abbreviations, anglicisms, neologisms, and emojis. At the same time, forums retain more traditional language forms. There is a phenomenon of hybridization of the language code, where elements of normative and situational speech are combined. The results also show that the language behavior of users is closely related to the technical features of the platform and the level of personalization of the information space. The practical significance lies in the possibility of using the results to update language policy in education, form digital language ethics, and support linguistic identity in the context of globalized communication. The data obtained can become the basis for further interdisciplinary research in the field of digital linguistics. In addition, the study makes a valuable contribution to IT research by demonstrating how digital platforms and algorithmic features directly shape linguistic behavior. This highlights the role of information technologies not only as neutral communication channels but also as active agents influencing cultural codes, which is significant for advancing interdisciplinary approaches at the intersection of linguistics and digital technology.
Keywords:	Digital Communication, Linguistic Norms, Language Transformation, Social Networks, Cultural Identity
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	DIGITAL STRATEGY ADAPTATION IN UNIVERSITY-BASED TECHNO PARKS: A SYSTEMATIC REVIEW OF INNOVATION AND ENTREPRENEURIAL ECOSYSTEMS
Author:	SIRILUK PHUENGROD, PINYAPHAT TASATANATTAKOOL, PRACHYANUN NILSOOK
Abstract:	This study presents a systematic literature review examining the strategic adaptation of digital technologies within university-based Techno Parks, with a particular focus on their role in fostering sustainable innovation and entrepreneurial ecosystems. Drawing upon 127 peer-reviewed articles published between 2019 and 2025, the review employs thematic synthesis guided by the PRISMA 2020 framework. The findings reveal that digital strategies particularly artificial intelligence, cloud computing, and big data analytics are increasingly central to the transformation of academic innovation environments. However, implementation challenges persist, including infrastructural limitations, digital skill gaps, and organizational inertia. Grounded in the Resource-Based View (RBV) and Dynamic Capabilities Theory, this study identifies key enablers such as agile governance, stakeholder collaboration, and absorptive capacity. By integrating digital transformation with sustainability imperatives, the study contributes to the theoretical discourse on innovation ecosystems and offers actionable insights for higher education leaders and policymakers. The review underscores the strategic potential of Techno Parks as digitally empowered platforms for advancing the United Nations Sustainable Development Goals (SDGs). Despite the growing literature on digital transformation in higher education, existing studies remain fragmented and largely technology-specific, with limited integration of theoretical frameworks such as the Resource-Based View (RBV) and Dynamic Capabilities Theory (DCT). This study addresses this gap by conducting a systematic literature review of 127 peer-reviewed articles published between 2019 and 2025. By synthesizing these findings through RBV and DCT, the review not only highlights prevailing digital strategies and associated challenges but also demonstrates how Techno Parks strategically adapt digital resources to advance sustainable innovation. The novelty of this work lies in its theoretical integration, its mapping of knowledge gaps in prior research, and its identification of new insights for higher education institutions, thereby providing a strong justification for the need and timeliness of this study.
Keywords:	Digital Transformation, Techno Parks, Higher Education Institutions, Innovation Ecosystems, Sustainable Development Goals
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	SECURING THE FUTURE: THE ENHANCED BB84 PROTOCOL FOR ROBUST QUANTUM KEY DISTRIBUTION
Author:	ARSHIYA MOBEEN M, SAFIA NAVEED S
Abstract:	Quantum Key Distribution (QKD), leveraging protocols such as BB84, offers information-theoretically secure key exchange essential for future communications. However, real-world IT deployment is compromised by practical limitations, including photon loss and hardware-based security flaws (e.g., non-ideal detectors enabling side-channel and Photon Number Splitting (PNS) attacks). The proposed Enhanced BB84 (E-BB84) protocol provides a significant contribution to information technology by providing an integrated framework to overcome these challenges and enable scalable quantum communication. E-BB84 purposefully incorporates state-of-the-art security countermeasures. Decoy states are used to mitigate PNS attacks and maintain high secret key rates. Measurement-device-independent QKD (MDI-QKD) eliminates all detector side-channel vulnerabilities, thus guaranteeing unconditional security. This protocol also uses entanglement-based techniques to improve the correlation and durability of communicators. To assess BB84 and E-BB84 performance under realistic eavesdropping and noise scenarios, a simulation framework was created using Qiskit. E-BB84's extensively better performance is established by way of comparative analysis that focuses on key performance indicators (Quantum Bit Error Rate, or QBER) and key rate. This highlights the capacity of E-BB84 to provide the necessary security and scalability required for practical quantum communication
Keywords:	Measurement-Device-Independent, Quantum Bit Error Rate, Quantum Key Distribution, Photon Number Splitting, Low-Density Parity-Check
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	SUPERGAN: SUPERPIXEL-GUIDED GAN FOR ROBUST AND IMPERCEPTIBLE IMAGE WATERMARKING
Author:	ABDELHAY HASSANI ALLAF, M HAMED AIT KBIR
Abstract:	This paper presents SuperGAN, a novel image watermarking framework that combines superpixel segmentation with a generative adversarial network (GAN) to achieve robust, imperceptible, and content-aware watermark embedding. By leveraging SLIC-based superpixel segmentation, regions with high texture complexity and sufficient size are selected for embedding to enhance both resilience and visual transparency. A 32-bit watermark is redundantly embedded across multiple regions to improve recovery under distortions. The embedding process integrates a region-aware attention mechanism to modulate watermark bits locally and imperceptibly. Extensive experiments on the COCO dataset assess the method’s performance under common attacks, including compression, noise, cropping, scaling, and rotation. Comparative results show notable improvements in robustness and imperceptibility. These findings highlight SuperGAN's ability to serve as a reliable tool for image watermarking applications in real-world.
Keywords:	Image Watermarking, Superpixels, Generative Adversarial Networks (GANs), Robustness, Attention Mechanism.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	UNCOVERING THE INTRINSIC NATURE OF TRAFFIC OSCILLATIONS: AN EMPIRICAL AND SIMULATION-BASED APPROACH
Author:	AHLAM EL ATTARI, KAMAL JETTO, ZINEB TAHIRI, ABDELILAH BENYOUSSEF, ABDALLAH EL KENZ
Abstract:	This paper challenges the conventional understanding of traffic oscillations, traditionally associated with congested conditions and minor disturbances. Through empirical analysis of diverse datasets, we demonstrate that traffic oscillations persist across various traffic states, conditions, and road section types. By applying curve fitting techniques to key traffic observables (space headway, time headway, and mean velocity), we quantified these oscillations and determined their characteristic periods. In free flow and jam states, the period corresponds to the average vehicle travel time (T₀), while in congested conditions, it equals 2πτ, where τ represents the relaxation time. To complement empirical findings, we employed Cellular Automata models, specifically the NaSch model, which allowed for parameter adjustment to simulate diverse traffic conditions. This approach corroborated empirical results and revealed that oscillation frequency in free flow is determined solely by maximum speed, while in jammed states, it depends exclusively on traffic density. These findings suggest that traffic oscillations are an intrinsic property of traffic flow dynamics, rather than merely a product of human behavior. Furthermore, this study establishes vehicle travel time as an effective metric for analyzing traffic flow in stable, homogeneous conditions.
Keywords:	Traffic Oscillation, Traffic Oscillations Period, Traffic Flow Dynamics, Empirical Analysis, Cellular Automata Models.
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text

Title:	AN ADAPTIVE GRAPH NEURAL NETWORKS WITH LANDSCAPE-AWARE PARTICLE SWARM OPTIMIZATION FOR INTELLIGENT MEDICAL INSURANCE FRAUD DETECTION
Author:	MR. V VINAY KUMAR, M V V A L SUNITHA, PALAMAKULA RAMESH BABU, A ARUNA KUMARI ARAJU ANITHA,BPRAVEENA MANDAPATI
Abstract:	This paper introduces a novel framework for detecting fraudulent medical insurance claims using Adaptive Graph Neural Networks (AGNN) optimized by Landscape-Aware Particle Swarm Optimization (LAPSO). By modeling healthcare data—including patients, providers, and procedures—as a graph, AGNN captures complex relational patterns indicative of fraud. Dynamic attention mechanisms help highlight critical relationships while LAPSO tunes key hyperparameters to improve generalization. Experimental results on real-world datasets demonstrate the proposed model’s superiority in terms of accuracy (91%), precision (89%), recall (88%), and F1-score (88%) over traditional machine learning and deep learning models. The results confirm the efficacy and interpretability of our framework for practical fraud detection applications. The AGNN component employs dynamic attention mechanisms to selectively prioritize significant relationships during message passing, thereby enhancing the ability of the model model’s ability to detect subtle and coordinated fraud schemes. To further improve detection accuracy and generalization, LAPSO fine-tunes critical hyperparameters of the AGNN architecture. By leveraging both global and local search capabilities, LAPSO accelerates convergence toward optimal configurations while adapting to the model’s performance landscape.The proposed method is evaluated on a real-world medical claim’s dataset, demonstrating superior performance associated with conventional deep learning and graph-based models among key metrics, including accuracy, precision, recall and F1-score. Moreover, the framework exhibits strong generalization capability, effectively identifying both known and previously unseen fraudulent behaviors. The integration of graph learning and evolutionary optimization offers a scalable and interpretable solution for healthcare providers and insurance companies aiming to mitigate fraud risks[15]. This study contributes to the progression of intelligent fraud detection systems and opens new directions for the application of adaptive graph learning in health. The AGNN component employs dynamic attention mechanisms to selectively prioritize significant relationships during message passing, improving the model’s capability to notice subtle and coordinated fraud schemes. To further improve detection accuracy and generalization, LAPSO fine-tunes critical hyperparameters of the AGNN architecture. By leveraging both global and local search capabilities, LAPSO accelerates convergence toward optimal configurations while adapting to the model’s performance landscape.
Keywords:	Adaptive Graph Neural Networks(Agnns),LAPSO,Deep Learning ,Medical Insurance,Health Care Analytics Etc.,
Source:	Journal of Theoretical and Applied Information Technology 15^th November 2025 -- Vol. 103. No. 21-- 2025
Full Text