Metaheuristic Optimization Review

AI-enabled digital twins are moving from simulation-oriented representations toward operational decision systems for critical infrastructure, but deployment evidence remains uneven and highly sector dependent. This review synthesizes AI-enabled digital twins in water and power infrastructure using a structured narrative evidence base of 36 primary application studies, balanced between 18 water and 18 power studies, together with recent reviews, interdependency research, and governance standards. The review compares application domains, data environments, AI roles, synchronization demands, validation settings, and decision authority across both sectors. Three analytical contributions extend beyond application cataloguing: a six-level deployment-maturity model with observable criteria; a coupled water–power framework for physical, cyber/data, spatial, and organizational cascading risk; and a trustworthy-AI pathway linking explainability, uncertainty, human oversight, fallback, traceability, interoperability, and re-validation. The evidence shows that water twins are strongly shaped by heterogeneous sensing, hydraulic/process uncertainty, and site-specific transfer, whereas power twins more often face tight latency, topology change, stability margins, and protection-adjacent decision requirements. Across both sectors, the dominant unresolved gap is not isolated AI accuracy but sustained field evidence demonstrating synchronized physical linkage, uncertainty-aware decisions, governed authority, cyber resilience, and lifecycle re-validation. The review therefore provides a cross-sector basis for distinguishing promising AI models from deployment-ready digital twins and for prioritizing future infrastructure-grade validation.

Modern communication networks have become increasingly heterogeneous, programmable, distributed, and servicedriven, creating management requirements that exceed the capabilities of purely manual or static rule-based operation. This review examines the evolution of intelligent Network Management Systems (NMSs) through the integration of machine learning, deep learning, predictive analytics, feature selection, metaheuristic optimization, network telemetry, software-defined control, and automated decision mechanisms. The reviewed methodological directions support traffic and performance prediction, anomaly and intrusion detection, fault localization, configuration optimization, resource allocation, routing assistance, quality-of-service management, capacity planning, and closedloop remediation. Intelligent approaches can strengthen situational awareness and reduce the time required to identify, diagnose, and respond to rapidly changing network conditions, while metaheuristic algorithms provide flexible mechanisms for feature-space reduction, model calibration, multi-objective resource optimization, and search within large configuration spaces. Practical deployment nevertheless remains constrained by telemetry quality, concept drift, computational overhead, interoperability, multi-vendor heterogeneity, security, privacy, explainability, scalability, and the risk of unsafe automated actions. Future research should therefore prioritize trustworthy, policy-aware, scalable, secure, and adaptive NMS architectures capable of combining continuous observability with verifiable closed-loop automation across enterprise, cloud, edge, Internet of Things, and multi-domain network environments.

Intelligent energy management is increasingly important for improving energy efficiency, reliability, economic performance, and sustainability across buildings, smart homes, industrial facilities, microgrids, and interconnected energy systems. This review examines recent advances integrating machine learning, deep learning, reinforcement learning, Internet of Things technologies, predictive analytics, feature selection, and metaheuristic optimization. These methods support energy-consumption forecasting, demand response, renewable-energy coordination, load scheduling, anomaly detection, building optimization, smart-home control, industrial energy management, and microgrid operation. The reviewed studies indicate that intelligent approaches can improve forecasting accuracy, reduce energy costs, optimize resource allocation, and enhance real-time decision-making under uncertain conditions. Metaheuristic algorithms further improve performance through parameter optimization, feature selection, and multiobjective scheduling. However, practical deployment remains challenged by data quality, computational complexity, cybersecurity, privacy, interoperability, generalizability, transparency, infrastructure cost, and communication reliability. Future research should prioritize explainable, scalable, secure, and adaptive intelligent energy-management frameworks for efficient, resilient, flexible, and low-carbon energy system operation.

Drug classification in clinical pharmacy is increasingly important for organizing medication knowledge, supporting therapeutic decisions, and improving patient-centered pharmaceutical care. As healthcare data become more complex, traditional rule-based classification approaches may be insufficient for interpreting heterogeneous drugrelated information, including medication histories, molecular descriptors, clinical indicators, adverse-event records, prescription patterns, and patient-specific risk factors. This review examines how artificial intelligence, machine learning, deep learning, feature selection, optimization, and decision-support systems contribute to drug classification in clinical pharmacy. The review discusses the role of intelligent models in medication safety assessment, drug–drug interaction monitoring, personalized therapy, pharmacy-service planning, and operational medication management. It also highlights major challenges related to data quality, interpretability, generalizability, workflow integration, and ethical implementation. Overall, intelligent drug classification can strengthen clinical pharmacy practice when computational models are transparent, clinically meaningful, and aligned with professional pharmaceutical judgment.

Diabetes prediction has become an important direction in clinical artificial intelligence because early identification of high-risk individuals can support preventive intervention, personalized monitoring, and improved long-term healthcare outcomes. This review examines recent machine learning, deep learning, and optimizer-assisted methodologies for diabetes prediction and related diagnostic decision support. It discusses how clinical variables, physiological measurements, biomedical signals, retinal images, population-specific attributes, and healthcare monitoring data can be transformed into predictive evidence through preprocessing, feature engineering, feature selection, model training, hyperparameter optimization, and risk interpretation. The review also emphasizes major methodological concerns, including class imbalance, model robustness, interpretability, subgroup variation, complication-oriented prediction, and deployment within connected healthcare environments. Overall, the reviewed literature indicates that effective diabetes prediction should be developed as an integrated clinical decision-support pipeline rather than as a single classifier comparison, with careful attention to data quality, feature relevance, model transparency, and practical healthcare usability.