3066-280XISSN (Online)
Brain-computer interface (BCI) systems based on electroencephalography (EEG) are applications that allow human-to-machine communication with intuitive (near-transparent) control, whose neural commands are decoded based on intentional movement. Recent research on the topic of machine learning (ML) has been able to greatly enhance the classification of the EEG-signals associated with the movement of the hands, head movements, and mobility movements of the eyes. The developments allow various utilization across assistive technologies, prosthetic control, and non-verbal communication. EEG, however, is highly non-stationery and noise-sensitive, so advanced preprocessing and optimization methods have to be applied to optimize performance in classification. This paper outlines an in-depth review of some of the most popular ML algorithms, i.e. support vector machines (SVMs), convolutional neural networks (CNNs), and recurrent neural networks (RNNs), and optimization methods, i.e., genetic algorithms (GAs), particle swarm optimization (PSO), and transfer learning. We point out existing problems in the processing of EEG signals and suggest directions in the future that will improve the robustness, generalization, and real-time behavior of BCI.
Read MoreDoi: https://doi.org/10.54216/MOR.050201
Vol. 5 Issue. 2 PP. 01-21, (2026)
Power quality disturbances (PQDs) have become an increasingly critical concern in modern power systems due to the rising integration of renewable energy resources, widespread use of power-electronic interfaced loads, and the growing complexity of cyber-physical grid infrastructures. These evolving conditions have introduced new sources of variability, uncertainty, and vulnerability into power networks, making it significantly more challenging to maintain voltage stability, waveform purity, and overall system reliability. Traditional deterministic methods for disturbance detection and classification are no longer sufficient to address the nonlinear, nonstationary, and high-dimensional nature of contemporary PQD phenomena. As the operational landscape grows more dynamic, there is a pressing need for analytical frameworks that can adaptively learn, generalize, and respond to diverse disturbance scenarios in real time. This study provides a comprehensive examination of recent advancements in PQD research, emphasizing the evolution toward hybrid analytical frameworks that integrate advanced signal processing, machine learning, and metaheuristic optimization. The literature demonstrates that hybrid models—such as continuous wavelet transform (CWT) combined with convolutional neural networks (CNNs), Stockwell transform integrated with kernel-based extreme learning machines (ELMs), and metaheuristic-optimized classifiers—significantly enhance detection accuracy, robustness against noise, and adaptability to varying operating conditions. These hybrid systems leverage the strengths of each component: signal transforms enrich the representational quality of PQD features, deep architectures facilitate automatic feature learning, and optimization algorithms refine model parameters to achieve optimal performance across complex and uncertain environments. Metaheuristic algorithms including Particle Swarm Optimization (PSO), Grey Wolf Optimization (GWO), Whale Optimization Algorithm (WOA), Differential Evolution (DE), and hybrid variants have proven particularly effective for feature selection, classifier optimization, system-level enhancement, and mitigation strategies. Their ability to handle large, multimodal, and nonlinear search spaces makes them especially suitable for modern PQD challenges, where disturbance signatures may overlap, evolve over time, or be obscured by noise introduced by renewable energy fluctuations. Furthermore, recent work demonstrates the potential of ensemble-based and deep learning models optimized with metaheuristics to outperform conventional approaches in both accuracy and computational efficiency, thereby advancing the state-of-the-art in PQD detection technologies. Additionally, the convergence of PQD analysis with cybersecurity highlights an emerging and increasingly urgent research frontier. As smart grids become more interconnected and reliant on information and communication technologies, they face heightened risks from cyber-attacks capable of inducing, masking, or mimicking PQ disturbances. Such adversarial actions pose significant threats to grid stability, integrity, and operational safety. Metaheuristic-enhanced deep learning methods have shown promise for cyber-physical intrusion detection by enabling classification models to identify subtle, intentionally disguised anomalies within PQ data streams. This hybrid approach provides a pathway toward resilient PQ monitoring frameworks that are capable of learning and adapting to evolving attack strategies. Despite notable advancements, several key challenges persist. First, the lack of standardized real-world datasets limits the generalizability and reproducibility of PQD research, particularly within renewable-dominated or cyber-physical grid environments. Second, the high computational demands of hybrid models hinder their deployment in real-time or resource-constrained settings, calling for advancements in lightweight architectures, model compression, and edge-intelligent PQD systems. Third, the field lacks unified frameworks that integrate PQ detection, classification, and mitigation with operational decision-making, economic constraints, and regulatory requirements. Finally, adaptive cyber-physical intrusion detection frameworks remain underdeveloped, especially for zero-day attacks and data-limited conditions. Overall, this review underscores the necessity of holistic, intelligent, and scalable approaches to PQD management. It identifies critical directions for future research, including real-time system integration, computationally efficient hybrid architectures, multiobjective optimization strategies, and robust cybersecurity-aware PQD analytics. These innovations are essential for achieving resilient next-generation smart grids capable of maintaining high power quality under increasingly dynamic, decentralized, and uncertain operating conditions.
Read MoreDoi: https://doi.org/10.54216/MOR.050202
Vol. 5 Issue. 2 PP. 22-33, (2026)
Oral cancer remains a significant global health concern, particularly due to the high rates of late-stage detection and the limitations of traditional diagnostic modalities. This study proposes a hybrid diagnostic framework that integrates deep learning with metaheuristic optimization to enhance the accuracy, efficiency, and interpretability of oral cancer classification. The architecture combines convolutional and recurrent neural network components with an adaptive optimization layer designed using swarm intelligence-inspired algorithms. This hybridization enables precise feature selection, architecture tuning, and parameter optimization, resulting in improved generalization and robustness across heterogeneous clinical datasets. The model is further augmented with explainable decision support features, enabling clinicians to visualize lesion relevance and interpret classification outcomes. Empirical evaluations demonstrate superior performance in terms of sensitivity, specificity, and computational efficiency compared to conventional training strategies. Additionally, the proposed framework is designed for portability and scalability, supporting potential deployment in mobile and edge-based diagnostic systems. The integration of interpretability, fairness constraints, and clinical adaptability underscores the model’s readiness for real-world implementation. This work contributes to the growing field of intelligent medical diagnostics and highlights the transformative potential of metaheuristic optimization in addressing complex, high-dimensional clinical classification tasks.
Read MoreDoi: https://doi.org/10.54216/MOR.050203
Vol. 5 Issue. 2 PP. 34-52, (2026)
This comprehensive review examines the current state of artificial intelligence and computational optimization techniques applied to earthquake engineering challenges. The paper systematically analyzes recent advances across three primary domains: machine learning (ML), deep learning (DL), and optimization methods, each contributing distinct capabilities to seismic hazard mitigation. Through an extensive analysis of peer-reviewed studies, this review synthesizes methodologies employed in earthquake prediction, early warning systems, structural damage assessment, emergency response optimization, and seismic hazard analysis. Machine learning approaches have demonstrated significant effectiveness in liquefaction prediction, slope displacement analysis, and seismic event classification, with models such as XG Boost and Random Forest achieving high predictive accuracy. Deep learning architectures, including convolutional neural networks (CNNs), recurrent neural networks (RNNs), and transformer-based models, have revolutionized real-time earthquake detection, P-wave recognition, and landslide susceptibility mapping, with several models achieving accuracy rates exceeding 90%. Optimization techniques, particularly metaheuristic algorithms like Particle Swarm Optimization (PSO) and Gray Wolf Optimizer (GWO), have proven valuable for emergency logistics, shelter allocation, and structural design optimization. The review reveals current trends toward hybrid frameworks integrating multiple computational approaches, enhanced model interpretability, and real-time implementation capabilities. Future research directions emphasize the development of uncertainty-aware models, scalable frameworks for global application, and integration of social and economic factors in disaster preparedness strategies. This review provides researchers and practitioners with a structured understanding of computational methodologies in earthquake engineering and identifies critical gaps requiring further investigation.
Read MoreDoi: https://doi.org/10.54216/MOR.050204
Vol. 5 Issue. 2 PP. 53-91, (2026)
Rainfall detection and forecasting are complex tasks in hydrology due to the nonlinear and multi-scale nature of precipitation processes. Recent advances in artificial intelligence (AI), deep learning, and metaheuristic optimization have significantly improved predictive accuracy across diverse geographic and climatic conditions. Deep learning models, such as ConvLSTM and hybrid CNN–LSTM systems, excel in capturing spatial and temporal dependencies, especially when combined with optimization algorithms like Whale Optimization and Ant Colony Optimization. These techniques help fine-tune model parameters, reduce errors, and prevent premature convergence. The integration of Extended Reality (XR) technologies, including Augmented Reality (AR) and Virtual Reality (VR), with AI-driven rainfall forecasting offers new opportunities for immersive visualization in water resource management. XR technologies enable real-time, interactive simulations of rainfall predictions and water distribution, enhancing decision-making for water management and climate adaptation planning. Despite challenges such as data scarcity and computational demands, the convergence of AI, metaheuristics, and XR technologies holds great promise for building resilient, accurate, and interpretable systems for global water resource management and flood mitigation.
Read MoreDoi: https://doi.org/10.54216/MOR.050205
Vol. 5 Issue. 2 PP. 92-109, (2026)
The increasing frequency and severity of seismic events worldwide demand innovative and adaptive solutions in earthquake engineering, early warning, and emergency response systems. Traditional deterministic optimization techniques often fall short in addressing the high-dimensional, nonlinear, and data-uncertain nature of many seismic problems. In contrast, metaheuristic algorithms—stochastic, population-based search methods inspired by natural phenomena—have emerged as powerful alternatives capable of providing robust and near-optimal solutions in complex environments. This review synthesizes the growing body of research on the application of metaheuristic optimization techniques across diverse earthquake-related domains. We examine over fifty influential studies that employ algorithms such as Genetic Algorithms (GA), Particle Swarm Optimization (PSO), Differential Evolution (DE), Ant Colony Optimization (ACO), Grey Wolf Optimization (GWO), and modern hybrid and multi-objective approaches. Applications span a wide spectrum—from seismic source localization and structural design, to tuned mass damper configuration, sensor placement, earthquake classification, and real-time emergency resource allocation. The review identifies key trends, including the evolution from single-algorithm methods to hybrid models that combine the strengths of multiple metaheuristics, and the transition from static to dynamic, real-time optimization frameworks. Addi-tionally, the integration of machine learning and reinforcement learning with metaheuristic search is shown to significantly improve the adaptability, accuracy, and performance of seismic systems. For instance, PSO-optimized neural networks and GA-tuned support vector machines have demonstrated enhanced precision in peak ground acceleration prediction and seismic zone classification. Despite their advantages, metaheuristic techniques face several open challenges. These include scalability to large-scale problems, lack of standard benchmarks and datasets, computational expense in high-fidelity simulations, and limited transparency in multi-stage or learning-augmented models. Moreover, reproducibility and generalizability of results remain underdeveloped due to inconsistent reporting standards and proprietary data. This review highlights the need for community-driven initiatives to establish open datasets, reproducible benchmarking platforms, and standardized performance metrics. Future directions emphasize lightweight, adaptive algorithms capable of operating in real-time environments, as well as interpretable and sustainable optimization frameworks suit-able for deployment on embedded systems and edge devices. In summary, metaheuristic optimization holds immense promise for advancing earthquake resilience. Its continued development—through hybridization, integration with AI, and emphasis on transparency and real-world applicability—will be instrumental in shaping the next generation of intelligent seismic risk mitigation tools.
Read MoreDoi: https://doi.org/10.54216/MOR.050206
Vol. 5 Issue. 2 PP. 110-124, (2026)