Application of Metaheuristic Algorithms in Optimizing Big Data Machine Learning Systems

With the explosive growth of big data and the increasing complexity of machine learning systems, optimizing the performance of these systems has become one of the fundamental challenges in the field of artificial intelligence. Metaheuristic algorithms, due to their inherent ability to search complex spaces and find near-optimal solutions, are considered powerful tools for improving the efficiency of machine learning models in big data environments. This analytical review paper, by examining a collection of reputable studies, classifies optimization approaches based on metaheuristic algorithms in big data machine learning systems. The approaches under review include hyperparameter optimization, feature selection for dimensionality reduction of big data, and architecture design of deep learning models. The results of this review indicate that hybrid methods can significantly reduce the number of features (up to ۱۰%) while substantially increasing prediction accuracy (between ۱% and ۱۰%), offering performance beyond conventional methods in real-world applications. Key challenges include computational complexity of algorithms, difficulty in parameter tuning, overfitting issues, model generalization, and interpretability. Future research directions include developing more efficient hybrid algorithms, using reinforcement learning for dynamic parameter control, integration with large language models, multi-objective optimization, federated learning for privacy preservation, and applications in interdisciplinary fields such as genomics, drug discovery, and medical Internet of Things.

کلیدواژه ها:

Metaheuristic Algorithms ، Machine Learning ، Big Data ، Hyperparameter Optimization