Asian Journal of Research in Computer Science

Multimodal Emotion Recognition Using Explainable Hybrid CNN-Transformer Networks

Sat, 04 Jul 2026 00:00:00 +0000

Emotion recognition is central to affective computing because it enables intelligent systems to perceive and respond to human affective states in domains such as healthcare, education, customer service, and human–computer interaction. However, unimodal methods based solely on facial expressions, speech, or text often fail in real-world settings due to noise, ambiguity, occlusions, and limited contextual cues, while many multimodal systems rely on simplistic fusion and remain difficult to interpret. This study proposes an Explainable Hybrid CNN-Transformer Network (EHCTN) for multimodal emotion recognition that integrates visual, audio, and textual information while improving transparency. Facial frames and audio spectrograms/MFCC-based representations are encoded via CNNs to capture discriminative spatial and acoustic patterns, and text is embedded using a pre-trained BERT model to obtain contextual semantics. Modality-specific features are combined using an attention-based fusion mechanism that dynamically weights each modality to strengthen robustness under noisy or partially missing inputs, followed by Transformer layers to model long-range dependencies and cross-modal interactions; a softmax classifier predicts emotion categories (e.g., happiness, sadness, anger, fear, surprise, neutral). Explainability is incorporated using Grad-CAM to localise salient facial regions and SHAP to quantify influential features across modalities. Experiments on IEMOCAP, MELD, and CMU-MOSEI with a 70/15/15 train–validation–test split and augmentation, trained in PyTorch on NVIDIA GPUs using AdamW (learning rate 1e-4, batch size 32, 100 epochs, dropout 0.5), show that EHCTN outperforms CNN, Transformer, and CNN-LSTM baselines, achieving 87.9% accuracy, 87.3% precision, 86.9% recall, and 87.1% F1-score, with reported accuracy gains of 11.4%, 6.2%, and 4.7% over the respective baselines. Confusion-matrix analysis indicates strongest performance for the Neutral class (228 correct) and minor confusion between Sad and Angry. Grad-CAM and SHAP analyses confirm reliance on meaningful facial regions (eyes, eyebrows, mouth), speech cues (e.g., pitch variation, intensity), and salient words, supporting trustworthy deployment of robust, interpretable emotion-aware systems.

A Machine Learning Based Early Warning Framework for CKDu Risk Prediction Using Water Quality Data

Nipuni Narmada Jayamaha, Maheesha Dhashantha Silva — Mon, 06 Jul 2026 00:00:00 +0000

This paper proposes and tests an environmental water-quality screening proof-of-concept based on machine learning and routine physicochemical measurements. Although the framework is described as CKDu risk prediction, it should be interpreted strictly as an environmental proxy water-potability feature-screening tool and not as a clinical patient-modelling system. The study employed an open-source dataset of 3,276 water samples with nine variables: pH, hardness, total dissolved solids, chloramines, sulfate, conductivity, organic carbon, trihalomethanes and turbidity. Missing values were imputed using median values, and feature scaling was applied where required. A stratified 80:20 train-test split was performed using the original class distribution and a random seed of 42. Accuracy, balanced accuracy, macro F1-score, ROC-AUC and confusion-matrix analysis were used to evaluate five supervised machine-learning models. Random Forest achieved the highest test performance, with an accuracy of 0.659, balanced accuracy of 0.641, macro F1-score of 0.636 and ROC-AUC of 0.695. The selected model was embedded in a prototype interface that translates non-potability probabilities into Low, Moderate and High screening bands and provides input validation and user-friendly follow-up messages. The results indicate the technical feasibility of multivariate machine learning for water-potability classification and sample prioritisation. The dataset does not include CKDu patient records, clinical outcomes, patient exposure histories, geographic exposure information or locally collected samples from CKDu-endemic communities. Therefore, the framework should be regarded only as a proof-of-concept environmental water-screening tool, not as a validated CKDu prediction, clinical diagnostic or regulatory decision-making system.

Governance-centered Artificial Intelligence Models for Enhancing Cyber Resilience, Data Privacy, and Regulatory Compliance in U.S. Healthcare Systems

Tue, 07 Jul 2026 00:00:00 +0000

The accelerating adoption of artificial intelligence across United States healthcare systems has intensified concerns over cyber resilience, data privacy, and regulatory compliance, yet existing governance frameworks remain fragmented and difficult to compare. This study addressed that gap by developing a governance-centred multi-criteria evaluation model (a structured governance-scoring instrument rather than a machine-learning system) that evaluates and ranks established frameworks on a single quantitative scale. Twelve governance constructs were derived from framework analysis and a control coverage mapping against adversary tactics, then weighted by combining the analytic hierarchy process with the entropy weight method to balance expert judgement and data-driven objectivity. The integrated weights produced composite readiness indices, dimensional resilience, privacy, and compliance scores, and a five-level maturity classification. Robustness was confirmed through bootstrap resampling, sensitivity perturbation, and consistency testing, returning a consistency ratio of 0.058 and rank stability of 0.958, while principal component analysis retained 86.3 percent of variance (interpreted descriptively given the small number of frameworks, n = 6). All six frameworks exceeded the readiness threshold, achieving a mean readiness of 0.782, with ISO/IEC 27001 ranked highest at 0.869 and the only Adaptive classification. The analysis isolated transparency and human oversight as the weakest constructs. The resulting instrument offers healthcare decision makers a reproducible, transparent, and internally consistent basis for strengthening accountability and prioritising critical governance improvements.

CNN-RNN Hybrid Model for Predicting Agricultural Yield from Soil Physico-Chemical Parameters

Wed, 08 Jul 2026 00:00:00 +0000

Smart agriculture represents an important challenge for food security and the sustainable optimisation of agricultural production systems. However, accurate agricultural yield prediction remains complex because of the nonlinear interactions among soil physicochemical parameters, environmental conditions and crop development over time. In this context, this study proposes a hybrid deep learning model that combines convolutional neural networks (CNNs) and recurrent neural networks (RNNs/LSTMs) to improve agricultural yield prediction from soil physicochemical data.

The adopted methodology is based on an experimental approach that integrates agricultural data collection, time-series pre-processing, variable normalisation and the fusion of spatial and temporal features within a hybrid CNN-RNN architecture. The variables considered include soil pH, nitrogen, phosphorus, potassium, organic matter, moisture and selected climatic data. Model performance was evaluated using the root mean square error (RMSE), mean absolute error (MAE) and coefficient of determination (R²), and was compared with that of classical models, including linear regression, random forests and support vector machines (SVMs).

R²R² The dataset, comprising 202 agricultural observations, was divided using an 80% training and 20% testing strategy to ensure an independent evaluation of model performance. The results show that the CNN-RNN hybrid model outperforms the traditional approaches, with an RMSE of 0.032, an MAE of 0.0221 and a coefficient of determination (R²) of 0.9895, indicating improved generalisation and higher accuracy in predicting agricultural yields. The proposed model also enables the identification of the soil parameters most influential in agricultural productivity, particularly pH and NPK nutrients.

This research shows that integrating hybrid deep learning techniques into precision agriculture can support the development of intelligent agricultural decision-support systems. It also provides a basis for the future integration of satellite data, IoT sensors and advanced artificial intelligence architectures in sustainable agriculture.

AI-Powered Cyber Warfare and the Evolution of Zero Trust Security Architectures in Autonomous Networks

Wed, 08 Jul 2026 00:00:00 +0000

The increasing sophistication of intelligent cyber warfare, in which adversaries exploit artificial intelligence to automate reconnaissance, generate polymorphic malware, and conduct machine-speed attacks, has rendered conventional perimeter security and static Zero Trust implementations inadequate for autonomous and self-managing networks. This study addresses the absence of an integrated, adaptive architecture by designing and analytically evaluating an AI-driven adaptive Zero Trust framework that unifies behavioural analytics, federated intrusion detection, explainable trust scoring, and autonomous policy enforcement within a NIST-aligned model. Adopting a quantitative, experimental, and simulation-based design, the framework was evaluated using public benchmark datasets including CICIDS2017, UNSW NB15, and BoT IoT, with standardised preprocessing, balanced resampling, and stratified cross-validation. Eight classifiers were trained, among which gradient boosting achieved an accuracy and F1 score of approximately 0.9999 on the CICIDS2017 benchmark after leakage-prone identifier features were removed, while ensemble and convolutional models performed strongly. The dynamic trust engine, exercised on an illustrative cohort of ten simulated agents, enforced conservative session-level access decisions, and a simulated three-node federated learning configuration produced an aggregated F1 score of 0.9185, quantifying the privacy-performance trade-off under heterogeneous partitions. These findings support a coherent, identity-centric defence delivering continuous, explainable verification. The study contributes a conceptual architectural blueprint validated through simulation rather than an operationally deployed system, and sampled datasets and simulated agents constrain operational generalization, motivating future validation on live autonomous network testbeds.

A Hybrid Convolutional Neural Network Approach for Context-aware Fashion Recommendation

Sanduni Dewmini Rupasinghe, Maheesha Dhashantha Silva — Wed, 08 Jul 2026 00:00:00 +0000

This study aimed to develop a fashion recommendation system that classifies clothing items and recommends complementary outfit pieces based on user-selected style preferences, addressing the limitation of existing systems that rely solely on visual similarity without incorporating style-based personalisation. The study was carried out at the Department of Information and Communication Technology, Faculty of Humanities and Social Sciences, University of Sri Jayewardenepura, Sri Lanka, between 25^th July 2025 and 2^nd June 2026, and followed a quantitative experimental approach involving the design, development, and performance evaluation of a two-stage deep learning-based classification and recommendation system. The proposed system consists of two stages: classification and recommendation. In the first stage, the system classifies the query item using a hybrid convolutional neural network (CNN) model combining ResNet-50 and EfficientNetB0. In the second stage, Fashion CLIP and CLIP ViT-B/32 models retrieve complementary items, which are then filtered and re-ranked based on the user-selected style from four categories: Casual, Formal, Party, and Streetwear. The classification model achieved 94% accuracy on the main dataset and 89.17% on external validation, while the recommendation pipeline achieved a mean Precision@5 of 84.2% and Accuracy@5 of 94.4%. The proposed system achieved 100% style consistency compared with 56.2% for the baseline model. The proposed two-stage system combines item classification with style-aware recommendation and has practical potential for integration into fashion e-commerce platforms to enhance user experience and support cross-category sales.

Cybersecurity and Digital Learning in Higher Education: Emerging Threats, Protective Technologies, and Future Directions

Thu, 09 Jul 2026 00:00:00 +0000

Higher education institutions occupy a uniquely exposed position in the global cybersecurity landscape. The convergence of large-scale digital transformation, open-access network philosophies, diverse user populations, and repositories of sensitive personal and research data has made universities consistently attractive targets for a wide spectrum of cyber threats. This critical review synthesises peer-reviewed literature and authoritative institutional evidence published between 2018 and March 2026, examining the evolving threat landscape confronting higher education institutions, with particular attention to ransomware, phishing, data breaches, insider threats, and the emerging risks associated with generative artificial intelligence. Protective technologies including zero-trust architectures, AI-driven intrusion detection, multi-factor authentication, and cloud security frameworks are critically assessed in relation to their applicability within academic environments. The human dimension of cybersecurity receives substantial attention, with cybersecurity awareness, training effectiveness, and behavioural factors examined across student and staff populations. Governance, policy, and regulatory considerations are discussed alongside the specific vulnerabilities of digital learning platforms and learning management systems. Findings indicate that whilst technological solutions are advancing, the most persistent vulnerabilities in higher education institutions remain structural and human, demanding integrated responses that combine robust technology governance with sustained cultural change. The review concludes by identifying key future directions, including quantum-resilient cryptography, federated security models, and sector-wide intelligence sharing.

Distributed Application Architecture and Key Technologies: Formal Analysis and Evaluation Framework for Microservice Adoption

Fri, 10 Jul 2026 00:00:00 +0000

Microservice architecture (MSA) has emerged as a dominant paradigm for designing and deploying distributed software systems, offering granular scalability, independent deployability, and technological heterogeneity at the cost of substantially increased operational complexity. Despite its widespread industry adoption, a coherent formal analytical basis and a generalisable evaluation framework for guiding MSA adoption decisions remain underdeveloped in the academic literature. This review synthesises scholarly research published between 2014 and April 2026 to examine the foundational concepts underpinning MSA, the enabling technologies that make it tractable in practice, and the formal analysis and quality assessment approaches proposed to guide architectural decision-making. The key technologies reviewed include containerisation, container orchestration, service mesh frameworks, application programming interface gateways, event-driven messaging systems, and distributed observability tooling. The review further examines migration strategies from monolithic systems, decomposition methodologies grounded in domain-driven design, and the application of formal verification methods to distributed service systems. A multi-dimensional evaluation framework is proposed, integrating organisational readiness, technical capability, security posture, and quality attribute trade-off analysis to support structured MSA adoption decisions. The review identifies significant gaps regarding standardised readiness assessment tools, empirically validated decomposition criteria, and formal methods applicable at the architectural level — gaps that point towards a pressing need for integrated, evidence-based frameworks capable of guiding practitioners through the complex socio-technical challenges of MSA adoption.

Student Engagement in Technology-enhanced Learning Environments and IT Employability Skills Development: An Analytical Review

Hasanthie Y. Dahanayake, M. G. M. Johar, Jacquline Tham — Fri, 10 Jul 2026 00:00:00 +0000

The rapid advancement of digital technologies has transformed higher education and increased the need to develop graduates with the technical competencies and workplace readiness skills required by the Information Technology (IT) industry. As a strategic approach, Technology-Enhanced Learning (TEL) has emerged to support student learning and digital engagement. However, the existing literature remains inconsistent regarding the relationships among technology-enhanced learning environments, student engagement and IT employability skills. This analytical review examines the relationship between technology-enhanced learning environments, student engagement and IT employability skills. Relevant studies were identified through a structured literature search of major academic databases using predefined inclusion and exclusion criteria. The review synthesises literature on digital learning platforms, learning management systems, collaborative tools, virtual laboratories and related technology-supported practices in higher education. It also considers student engagement as a multidimensional construct comprising behavioural, emotional and cognitive engagement. The selected studies were analysed through thematic synthesis to compare findings, identify recurring themes and examine areas of inconsistency. The review indicates that effective technology-enhanced learning environments with strong system capability and high system quality can improve behavioural, emotional and cognitive engagement, which may help IT students develop technical skills and workplace readiness. The paper identifies gaps in existing research, including limited integrated models, insufficient attention to emotional and cognitive engagement, inadequate focus on IT-specific employability skills and limited evidence from developing-country contexts. The review contributes to the existing literature by synthesising available evidence and providing practical implications for educators, curriculum designers, higher education institutions and policymakers seeking to enhance graduate employability through technology-enhanced learning.

Hybrid Classical–Quantum-Inspired Neural Network with Simulated Variational Circuit for Credit Card Fraud Detection

V. R. Srividhya, Srividhya Ganesan — Fri, 03 Jul 2026 00:00:00 +0000

Credit card fraud detection remains a challenging binary classification problem because fraudulent transactions are rare, transaction patterns are complex, and false negatives may have important operational consequences. This study presents a Hybrid Classical–Quantum-Inspired Neural Network (HCQNN) with a simulated variational quantum circuit for credit card fraud detection. The proposed framework combines classical preprocessing, SMOTE-based class balancing, neural network-based feature learning, and quantum-inspired variational feature transformation. The model was evaluated using the Credit Card Fraud Detection dataset after applying SMOTE to the training data and was compared with three classical baseline classifiers: Logistic Regression, Decision Tree, and Linear Support Vector Machine. The experimental results show that the proposed HCQNN achieved an AUC of 97.85%, precision of 91.20%, recall of 90.10%, and an F1-score of 90.64%. These values indicate improved classification balance, particularly in the detection of minority-class fraud cases, compared with the selected baseline models. Training and validation behaviour also showed stable convergence, with training and validation accuracies exceeding 96% and 95%, respectively. Since the variational quantum circuit was simulated on classical hardware, the findings should be interpreted as evidence of the value of hybrid feature learning and quantum-inspired transformations rather than as proof of quantum computational advantage. The study provides a basis for further evaluation using broader datasets, additional baseline models, and real quantum hardware.