Browsing by Author "Raini Hassan"

Mental fatigue is one of the critical issues in this world that affects the overall well being of people in terms of cognitive performance, emotions, and decision making. Mental fatigue is one of the typical human infirmities. Studies report that sleep deprivation and long work hours increase the likelihood of fatigue and fatigue related accidents, and contribute to reduced productivity, errors, and accidents. Fatigue is one of the biggest causes of crashes involving cars, lorries, and buses. Traditionally, mental fatigue has been assessed through subjective methods such as interviews and psychometric questionnaires, which are prone to bias, inconsistency, and imprecision. Failure to address mental fatigue effectively may lead to impaired decision making, reduced safety in transportation and healthcare, and overall decline in quality of life. This research aims to address these gaps by proposing the Neuro Physiological Emotional Profiling Model for Mental Fatigue (NPEMMF), which integrates electroencephalography (EEG) data with human emotional stimuli to provide a more objective and reliable assessment of mental fatigue. Specifically, the objectives are: (i) to identify the relationship between mental fatigue and its consequences on human emotion, (ii) to develop a neurophysiological emotional profiling model for mental fatigue based on ERP features, and (iii) to evaluate the performance of the neurophysiological profiling model based on the affective space model for detecting the underlying emotions in mental fatigue. Event Related Potential (ERP) was chosen due to its sensitivity in capturing time locked brain responses to emotional stimuli, while the Wide Neural Network (WNN) was used as the classifier to analyze the gathered data to find trends and create a model for mental fatigue due to its robustness in handling nonlinear and high dimensional EEG data. The International Affective Picture System (IAPS) was used as a stimulus instrument for emotions such as happy, calm, fear, and sad. Experimental results indicated that the peak to peak data produced more reliable and consistent results compared to peak to peak and latency data. EEG channel analysis according to the affective space model revealed that for positive arousal, the frontal EEG channels of F3 and F4 are most appropriate for studying emotions happy and fear. On the other hand, the EEG channel Cz was suitable for studying emotions with negative arousal. The ERP components that are most suitable for positive valence are N1, P1, N2, P2, N3, and P3. For negative valence, the most suitable ERP components are the Late Positive Potentials (LPP). The evaluation confirmed that the NPEMMF framework reduces bias and inconsistency compared to traditional subjective approaches, providing a more objective and accurate profiling of mental fatigue. Overall, this research contributes to advancing knowledge in mental fatigue assessment and emotional profiling. The proposed framework has potential applications in high risk domains, such as the transportation sector, where reliable detection and regulation of mental fatigue can enhance safety and decision making.

In today`s world that continuously processes data, offline or online, data is accumulating every day, which create difficulties for the existing data processing, such as classification to catch up. The more the data means, the more it requires time for processing, and may cause data overfitting, and this will conflict with today`s lifestyle that demands faster and accurate results. Many researchers in this area are focusing on applying Feature Selection (FS) techniques that will reduce the number of features. However, based on the reviews, none is working together with Input Significance Analysis (ISA) techniques, which can provide meaning for each feature in the dataset before being processed by the classifiers. Additionally, ISA can offer some insights about the “black box” element inherited by the classifiers; that hides any details about the classification processes and results derivation, which later can trigger doubts and questions on how such classification results produced. The methodology of this research comprises of six groups of experiments or stages. In the first three stages, the feature ranking method is performed, as part of ISA implementation. The last three stages performed the feature selection, as part of FS implementation. The preliminary results, obtained from the first three stages, showed that the percentage of error rate is decreasing by using ranked dataset. From the last three stages, as final results, the ranked dataset with feature selection has been found to produce improved results compared to the original and complete dataset. In summary, after the original and complete dataset has been interpreted well by ISA, together with the implementation of FS that reduce the number of features according to the weights obtained and ordered by ISA, training has become faster, the size of the network has been reduced, and more accurate results has been produced.

Skyline queries play a critical role in multi-criteria decision-making by identifying data points that are not dominated by others, thus offering optimal choices to users. These queries are particularly valuable in transportation management, logistics, route optimization, and decision support systems. However, existing skyline query processing algorithms exhibit limited effectiveness when applied to large-scale and uncertain graph datasets, due to their reliance on exhaustive dominance comparisons, and sensitivity to uncertainty, which collectively result in incompatibility when applied in real-world graph environments. To address these challenges, this research proposes a novel skyline query processing framework that integrates Graph Neural Networks (GNNs) with Reinforcement Learning (RL), enabling effective representation learning over uncertain graph structures and adaptive skyline selection, a solution to ensure compatibility, and also test the potential scalability of the framework. As part of the contribution, large-scale uncertain graph datasets are systematically constructed with controlled size, density, and uncertainty levels to enable rigorous evaluation and scalability analysis. The proposed method is evaluated using 10-fold cross-validation, and performance is measured using accuracy, precision, recall, F1-score, and ROC-AUC. Experimental results demonstrate that while baseline skyline algorithms achieve acceptable accuracy, they suffer from significantly lower precision and recall, leading to suboptimal identification of skyline points. In contrast, the proposed GNN-RL framework achieves an accuracy of 98.97% alongside recall and F1-score above 98%, demonstrating strong robustness in uncertain graph settings. Furthermore, scalability experiments across varying dataset sizes confirm the suitability of the proposed approach for large-scale skyline query processing. This research contributes both theoretically and practically to intelligent data analytics and supports the United Nations Sustainable Development Goal (SDG) No. 9 which promotes resilient infrastructure, sustainable industrialization, and innovation through the development of scalable and intelligent data-driven technologies.

Skyline queries are widely used in multi-criteria decision-making to identify non-dominated data points which balance conflicting preferences. While skyline computation has been studied extensively in relational and complete databases, limited attention has been given to skyline query processing in large-scale incomplete graph databases. The challenges include the dynamic and evolving nature of graphs with frequent additions and deletions of nodes, the prevalence of missing attribute values that disrupt dominance relationships and reduce query reliability. These challenges become critical in real-world applications such as recommendation systems, urban planning, fraud detection and location-based services, where incomplete or sparse data is common. Traditional approaches relying on relational-to-graph transformation and heavy preprocessing suffer from inefficiency, sparsity, and poor scalability when applied to high-dimensional graph data. The proposed study introduces an optimized framework for skyline query processing in incomplete graph databases by integrating machine learning techniques, including clustering-based optimization with the K-Means algorithm, dynamic data pruning, and adaptive indexing. The framework reduces computational overhead, handles missing values more effectively, and ensures accurate skyline retrieval under incomplete graph database. Experimental evaluation on synthetic graph datasets, designed to real-world incompleteness, demonstrates the framework's effectiveness. The proposed method achieves a reduction in query processing time of 30-50% and a dataset size reduction of up to 44.44% compared to traditional baseline algorithms. Cluster quality was validated using intrinsic metrics such as the Silhouette Score, ensuring the robustness of the groupings. The proposed solution significantly advances skyline query processing for complex, incomplete graph structures, contributing to more efficient and reliable decision-support systems, recommendation engines, and location-based services.

Skyline query processing is essential in multi-criteria decision-making, as it retrieves optimal results without requiring user-defined weights. Traditional skyline methods, however, face significant challenges when applied to large-scale and incomplete datasets. This study proposes a hybrid approach that integrates the ISkyline dominance graph technique with Graph Neural Networks (GNNs), specifically a Graph Convolutional Network (GCN) to improve skyline query performance under such conditions. The GCN component is utilized to predict skyline tuples in the presence of missing or incomplete data. The ISkyline algorithm serves as the foundation for identifying initial dominance relationships and labelling skyline points, enabling the GCN to learn Pareto-optimal patterns from partially incomplete data. Evaluation on both synthetic and real-world datasets demonstrates enhanced accuracy and efficiency when compared to established methods such as ISkyline, SIDS, and OIS. The proposed GNN + ISkyline framework improved classification accuracy by 72%, the F1-score by 71%, and the AUC-ROC by 49% compared to the standalone ISkyline algorithm when evaluated on the CoIL 2000 dataset. This work demonstrates the potential of creating a more efficient query processing, supporting applications in e-commerce, finance, and smart data systems, while aligning with the 9th Sustainable Development Goal on industry, innovation, and infrastructure.