Garuda - Garba Rujukan Digital

Interpretable Academic Outcome Prediction Using Explainable Boosting Machines Godfrey Perfectson Oise; Felix Oshiorenoya Uloko; Kevin Chinedu Pius; Enovwo Eferoba–Idio; Michael Uyiosa Edobor; Evans Mintah; Osahon Ukpebor; Oludare Sokoya; Tejiri Jessa
Methods in Science and Technology Studies Vol. 2 No. 1 (2026): June
Publisher : PT. Teknologi Futuristik Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.64539/msts.v2i1.2026.441

Predictive analytics has become an important component of learning analytics in higher education, enabling institutions to identify academic risks and support student success through data-driven decision making. However, many existing academic outcome prediction models rely on complex black-box machine learning techniques that provide high predictive performance but limited transparency and interpretability. The lack of explainability restricts the practical adoption of such models in educational environments where accountability, trust, and ethical decision-making are essential. This study proposes an interpretable machine learning framework for multi-class academic outcome prediction using the Explainable Boosting Machine (EBM), a glass-box model that combines the predictive power of ensemble boosting with the transparency of generalized additive models. The proposed framework was evaluated using a publicly available Student Performance and Learning Behavior dataset consisting of 6,519 student records containing academic, behavioral, and demographic attributes. Academic outcomes were formulated as a four-class classification task: Distinction, Pass, Fail, and Withdrawn. Model performance was assessed using multiple evaluation metrics including accuracy, precision, recall, F1-score, and ROC–AUC analysis. Experimental results demonstrate that the proposed EBM model achieves an accuracy of 88% and an overall ROC–AUC score of 0.91, indicating strong predictive capability across outcome categories. Furthermore, the model provides native interpretability through feature contribution functions and SHAP-based explanations, enabling both global and instance-level understanding of predictions. The results demonstrate that reliable academic outcome prediction can be achieved without sacrificing interpretability, providing a transparent and trustworthy decision-support framework for educational analytics.

Deep Learning–Driven Anomaly Detection for IoT-Enabled Smart Engineering Systems Godfrey Perfectson Oise; Kevin Chinedu Pius; Felix Oshiorenoya Uloko; Immunhierokene Clinton Obrorindo; Roli Lydia Oshasha
Scientific Journal of Engineering Research Vol. 2 No. 3 (2026): September
Publisher : PT. Teknologi Futuristik Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.64539/sjer.v2i3.2026.432

The rapid adoption of Internet of Things (IoT) technologies in smart engineering systems has increased the need for reliable anomaly detection mechanisms capable of identifying cyberattacks, operational faults, and abnormal system behaviors in complex cyber–physical environments. Existing rule-based and conventional machine learning approaches often struggle to effectively model the non-linear, high-dimensional, and highly imbalanced nature of IoT-generated multivariate time-series data, thereby limiting their capability to detect subtle and previously unseen anomalies. To address these challenges, this study proposes a deep learning–driven anomaly detection framework based on a hybrid CNN–LSTM autoencoder architecture for modeling spatiotemporal system behavior in IoT-enabled engineering environments. The proposed framework integrates convolutional neural networks for spatial feature extraction with long short-term memory networks for temporal dependency learning, while anomaly detection is performed using reconstruction error analysis and adaptive thresholding under unsupervised learning conditions. Experimental evaluation was conducted using the BATADAL-A dataset, which represents a realistic cyber–physical water distribution system. The results demonstrate stable convergence and strong generalization performance, with closely aligned training and validation losses throughout the learning process. The proposed framework achieved 90% overall accuracy, anomaly precision of 0.83, anomaly recall of 0.22, and an AUC of 0.677, indicating effective modeling of normal operational behavior but limited sensitivity to rare anomalous events. These findings demonstrate that the proposed CNN–LSTM autoencoder provides reliable low–false alarm monitoring for IoT-enabled smart engineering systems while highlighting the need for future improvements to enhance anomaly sensitivity and robustness in safety-critical applications.

Energy-Efficient Federated Learning with Temporal Convolutional Networks for Intrusion Detection Godfrey Perfectson Oise; Felix Oshiorenoya Uloko; Kevin Chinedu Pius; Roli Lydia Oshasha; Eric Edeigue Osemwegie; Immunhierokene Clinton Obrorindo
Methods in Science and Technology Studies Vol. 2 No. 1 (2026): June
Publisher : PT. Teknologi Futuristik Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.64539/msts.v2i1.2026.462

The rapid proliferation of Internet of Things (IoT) devices has significantly increased the attack surface of modern network infrastructures, necessitating intelligent and scalable intrusion detection systems. Federated Learning (FL) has emerged as a promising paradigm for distributed model training without centralized data sharing; however, challenges such as energy efficiency, data heterogeneity, and privacy preservation remain inadequately addressed. Existing studies often emphasize optimization objectives theoretically without validating them under realistic constraints. This paper proposes an energy-aware federated learning framework integrating Temporal Convolutional Networks (TCNs) for intrusion detection using distributed network traffic data. The framework incorporates differential privacy for secure model updates and a conceptual energy-aware client participation strategy. Experiments are conducted on the UNSW-NB15 dataset under a controlled setting with fixed client participation and communication parameters. The results demonstrate that the proposed model achieves improved classification accuracy and stable convergence behavior across communication rounds while operating under a fixed energy budget. However, energy consumption remains constant due to controlled experimental conditions, indicating that the study evaluates performance under energy constraints rather than dynamic energy optimization. The findings highlight the effectiveness of TCN-based federated models for intrusion detection in resource-constrained environments. Future work will focus on dynamic energy modeling, heterogeneous client environments, and comprehensive multi-objective evaluation.

A Convolutional Neural Network Framework for Intelligent Intrusion Detection Godfrey Perfectson Oise; Babatunde Seyi Olanrewaju; Oshoiribhor Austin Orukpe; Kevin Chinedu Pius; Augustine Osazee Airhiavbere
Scientific Journal of Computer Science Vol. 2 No. 1 (2026): June
Publisher : PT. Teknologi Futuristik Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.64539/sjcs.v2i1.2026.404

The rapid expansion of cloud computing, Internet of Things (IoT), and distributed network environments has significantly increased vulnerability to sophisticated cyber threats, exposing the limitations of traditional signature-based intrusion detection systems. Although deep learning techniques, particularly Convolutional Neural Networks (CNNs), have shown promising performance in intrusion detection, challenges related to validation transparency, statistical reliability, and interpretability remain inadequately addressed. This study proposes an intelligent CNN-based intrusion detection framework designed to improve detection accuracy, robustness, and model explainability. The framework is evaluated using the UNSW-NB15 benchmark dataset, which reflects realistic modern cyber-attack scenarios. A comprehensive preprocessing pipeline involving data cleaning, categorical encoding, feature normalization, and data reshaping is applied to enhance learning efficiency. To ensure unbiased evaluation, stratified k-fold cross-validation and an independent held-out test set are employed. Experimental results demonstrate that the proposed CNN achieves a test accuracy of 91.8%, with balanced precision, recall, and F1-score across benign and malicious traffic classes. Multi-class detection analysis further confirms the model’s capability to distinguish among diverse attack categories. Statistical validation using mean performance metrics, standard deviation, and confidence intervals demonstrates stable generalization performance. Additionally, Gradient-weighted Class Activation Mapping (Grad-CAM) is used to enhance interpretability by identifying network-level features that influence classification decisions. An ablation study further validates the effectiveness of key architectural components. The results indicate that the proposed framework provides a reliable, scalable, and interpretable solution for intelligent intrusion detection in modern high-dimensional network environments.

Isolation Forest–Based Intrusion Detection for Cyber-Physical Systems Godfrey Perfectson Oise; Susan Konyeha; Felix Oshiorenoya Uloko; Kevin Chinedu Pius; Enovwo Eferoba–Idio; Michael Uyiosa Edobor; Evans Mintah; Osahon Ukpebor; Oludare Sokoya; Tejiri Jessa
Scientific Journal of Engineering Research Vol. 2 No. 2 (2026): June
Publisher : PT. Teknologi Futuristik Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.64539/sjer.v2i2.2026.434

Cyber-physical engineering systems (CPES) form the backbone of critical infrastructures such as power generation, industrial automation, and water treatment facilities. Because cyber intrusions in these environments can directly disrupt physical processes, reliable intrusion detection mechanisms are essential for maintaining operational safety and system resilience. However, many existing intrusion detection approaches rely on supervised learning techniques that require large volumes of labeled attack data, which are rarely available in real industrial environments. In addition, advanced detection methods often introduce significant computational overhead, limiting their practicality for deployment in resource-constrained cyber-physical systems. To address these challenges, this study proposes a one-class anomaly detection framework based on the Isolation Forest algorithm for monitoring cyber-physical engineering systems. The proposed approach learns the statistical distribution of normal operational behavior using multivariate sensor, actuator, and control signals, and identifies deviations from this learned pattern as potential cyber intrusions. The framework is evaluated using the Hardware-in-the-Loop–based Augmented Industrial Control System (HAI) Security Dataset, which provides realistic industrial process measurements under both normal and attack scenarios. Experimental results show that the model achieves overall accuracy (0.89) and strong performance in identifying normal operational states (F1-score = 0.94). However, attack detection shows moderate recall (0.48) but low precision (0.04) due to class imbalance and overlapping anomaly score distributions. These findings indicate that Isolation Forest serves as a computationally efficient baseline anomaly detection mechanism for real-time CPS monitoring, while highlighting the need for hybrid and temporally aware detection strategies to improve attack discrimination in industrial cyber-physical environments.

Facial Expression Recognition Using a Sequential Convolutional Neural Network for Multi-Class Classification Godfrey Oise; Immunhierokene Clinton OBRORINDO; Roli Lydia OSHASHA; Kevin Chinedu PIUS; Felix Oshiorenoya ULOKO
Indonesian Journal of Modern Science and Technology Vol. 2 No. 1 (2026): January
Publisher : CV. Abhinaya Indo Group

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.64021/

Facial Emotion Recognition (FER) has become an important area of research in affective computing, human–computer interaction, intelligent surveillance, and healthcare applications due to its ability to automatically identify and interpret human emotional states from facial expressions. This study presents a lightweight Sequential Convolutional Neural Network (S-CNN) framework for multi-class facial emotion recognition using facial expression images categorized into eight emotional classes: Anger, Contempt, Disgust, Fear, Happy, Neutral, Sad, and Surprised. The proposed framework integrates image preprocessing, data augmentation, convolutional feature extraction, and deep learning-based classification to develop an efficient and computationally lightweight emotion recognition system. The model was trained and evaluated using a publicly available facial expression dataset, with performance assessed using accuracy, precision, recall, F1-score, and confusion matrix analysis. Experimental results demonstrated strong classification performance, achieving an overall accuracy of 92%, a weighted precision of 87%, a weighted recall of 92%, and a weighted F1-score of 89%. The confusion matrix further revealed effective discrimination among most emotional categories, with only minor misclassification observed between visually similar expressions. Comparative analysis with established deep learning architectures reported in the literature, including VGG16, MobileNetV2, ResNet50, and EfficientNet-B0, highlights the potential effectiveness of the proposed lightweight architecture while maintaining lower computational complexity. The findings demonstrate that simplified CNN architectures can provide accurate and efficient facial emotion recognition, making them suitable for real-time and resource-constrained applications. Future research should explore larger benchmark datasets, cross-dataset validation, and advanced deep learning architectures to further improve generalization and robustness.

Title

Found 6 Documents
Search

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Title Search

Found 6 Documents Search

Abstract

Abstract

Abstract

Abstract

Abstract

Abstract

Title

Found 6 Documents
Search