Metaheuristic Optimization of Deep Learning Models for Early Diabetes Prediction: Current Trends, Applications, and Future Perspectives

The escalating global prevalence of diabetes mellitus has intensified the demand for highly accurate predictive models. Deep learning architectures, particularly Convolutional Neural Networks (CNNs), have demonstrated superior performance in diabetes risk assessment using clinical, laboratory, and lifestyle data. However, their effectiveness is heavily dependent on optimal hyperparameter configuration, a complex and computationally expensive task. Metaheuristic algorithms such as Genetic Algorithms (GA), Particle Swarm Optimization (PSO), Grey Wolf Optimizer (GWO), Simulated Annealing (SA), Differential Evolution (DE), and Ant Colony Optimization (ACO) have emerged as powerful tools for automated and efficient hyperparameter tuning and feature selection in deep learning models. This systematic review analyzes ۳۰ high-impact studies published between ۲۰۱۶ and ۲۰۲۵, categorizing metaheuristic-deep learning hybrid frameworks, evaluating their impact on predictive performance, and identifying current limitations. Results consistently show accuracy improvements of ۴–۱۲% over grid/random search baselines, with hybrid GA-PSO-CNN and GWO-CNN models achieving state-of-the-art results (>۹۶% accuracy) on benchmark datasets. The review concludes with a critical roadmap for clinical deployment and future research directions.

کلیدواژه ها:

Diabetes Prediction ، Deep Learning ، Convolutional Neural Networks ، Metaheuristic Optimization ، Genetic Algorithm ، Particle Swarm Optimization ، Hyperparameter Tuning ، Feature Selection ، Hybrid Frameworks