A team of researchers from the University of Professional Studies, Accra, led by Dr Augustina Dede Agor, has completed one of the most detailed global reviews to date on the integration of artificial intelligence and optimization algorithms.
The study, which is forthcoming in the Journal of Artificial Intelligence and Technology (JAIT), published by Intelligence Science and Technology Press Inc., investigates how machine learning (ML) is being combined with metaheuristics (MH) as well as how different metaheuristics are being hybridized with one another to solve increasingly complex real-world problems. Covering research published between January and October 2024, this work offers a rare month-by-month and country-level analysis of trends shaping global computational intelligence research.
The research adopts a bibliometric-scoping methodology, utilizing structured, logic-based searches of Google Scholar and a PRISMA-guided screening process to identify 119 peer-reviewed publications that explicitly introduce new hybrid metaheuristic–machine learning (MH-ML) or metaheuristic–metaheuristic (MH-MH) algorithms.
In total, 14 MH-ML and 105 MH-MH techniques were examined. Each paper was reviewed manually to classify algorithm combinations, domain applications, and publication sources, providing a more nuanced perspective than automated bibliometric tools alone.
The findings reveal a striking difference in maturity between the two approaches. MH-MH research has shown broad and stable growth, dominating the field with steady monthly publication rates and strong representation across journals and countries.
India emerged as the world leader in this category, publishing 46 of the 105 MH-MH studies, with a strong focus on energy forecasting, industrial job scheduling, and urban logistics optimization- areas aligned with its national digitalization agenda. China followed with 15 contributions, while countries such as Iran, Iraq, Turkey, and Egypt have also become active players, demonstrating a diversification of expertise beyond traditional Western research centers.
By contrast, MH-ML research, although rapidly growing, remains in an early stage of development. China again leads this space, contributing five of the 14 MH-ML studies, with Saudi Arabia closely following. Several of these studies highlight the integration of reinforcement learning, deep neural networks, and clustering algorithms into classical metaheuristics to improve decision-making, convergence, and computational efficiency.
Notable innovations include hybrid frameworks such as a Deep Q-Network-driven Memetic Algorithm for flexible job shop scheduling, an Enhanced Seagull Optimization Algorithm combined with deep reinforcement learning for intrusion detection in IoT networks, and the integration of graph neural networks with genetic algorithms for traffic assignment modeling in road network design.
The analysis also shows a clear preference for certain techniques within these hybrids. In machine learning, Convolutional Neural Networks (CNNs) and K-means clustering emerged as the dominant approaches. CNN’s strength in hierarchical feature extraction and its versatility in image and time-series data make it a preferred choice for energy consumption forecasting, IoT security, and robotics applications.
K-means clustering is often used to initialize or guide search operators within metaheuristics, improving their ability to escape local optima. In the realm of metaheuristics, Genetic Algorithms (GA) were the most widely used in MH-ML integrations, often combined with neural networks to optimize complex model parameters. Particle Swarm Optimization (PSO) has dominated MH-MH research, reflecting its enduring popularity due to its fast convergence and ability to handle high-dimensional optimization problems with relatively low parameter sensitivity.
The review further identified significant publication trends across major journals and publishers. Elsevier was the most active publisher in ML-driven hybrid research, with journals such as Expert Systems with Applications and Computers & Industrial Engineering featuring several high-impact studies.
Expert Systems with Applications has become a leading venue for research that combines AI with practical optimization scenarios, while Computers & Industrial Engineering focuses on predictive modeling, resource scheduling, and decision support- topics that mirror the hybrid solutions being developed. For MH-MH research, Springer’s Cluster Computing journal emerged as the most prominent platform, featuring multiple studies on theory-driven metaheuristic innovation and benchmark testing.
A closer examination of application domains underscores the practical significance of these techniques. Hybrid algorithms are being increasingly deployed in smart energy systems, with several studies focusing on predictive modeling of consumption patterns in public buildings and renewable energy networks.
Industrial scheduling, particularly for flexible job shop environments that integrate autonomous guided vehicles, has also benefited from hybrid designs that optimize both manufacturing and logistics simultaneously. Other impactful applications include autonomous rescue robotics, cybersecurity, wireless sensor networks, and intelligent traffic control, fields where optimization speed and adaptability are critical.
This forthcoming study is not only a snapshot of the state of hybrid optimization in 2024 but also a roadmap for its future development. The research reveals that reinforcement learning remains underutilized in hybrid optimization, an insight previously echoed in the literature, but here demonstrated with precise metrics. The growing participation of countries in the Middle East and Africa, alongside established hubs in China and India, signals a significant shift toward a more geographically distributed innovation ecosystem.
Dr. Agor emphasizes that this work demonstrates Ghana’s leadership in computational intelligence research, stating:
“Algorithm innovation is no longer confined to isolated research groups. Machine learning and metaheuristics are converging into highly adaptive, globally significant systems. This review not only consolidates these advances but provides actionable intelligence for engineers, scientists, and policymakers developing optimization-driven solutions for energy systems, logistics, and security.”
By documenting publication trends, algorithm preferences, and research collaborations, this forthcoming article in the Journal of Artificial Intelligence and Technology (JAIT), published by Intelligence Science and Technology Press Inc., establishes a foundation for future exploration of hybrid AI-driven optimization strategies. It represents a crucial step toward developing more efficient, adaptive, and sustainable computational systems that address the growing complexity of global challenges.
About the Author
Dr. Augustina Dede Agor is a lecturer in the Department of Information Technology at the University of Professional Studies, Accra. She holds a PhD in Computer Science and has over 11 years of experience in academia and research.
Her expertise encompasses artificial intelligence, computational optimization, and metaheuristics, as well as neural networks and unsupervised learning techniques such as Kohonen Self-Organizing Maps (SOMs).
She also specializes in biometric identification systems, including Automated Fingerprint Identification Systems (AFIS), and conducts research in computer networks and communication systems.
By Dr. Augustina Dede Agor, Department of Information Technology, University of Professional Studies, Accra.
Contact: augustinadede.agor@upsamail.edu.gh
Copyright © 1994 - 2025 GhanaWeb. All rights reserved.
