Application of a Hybrid Metaheuristic Optimization Algorithm for Feature Selection to Improve Classification Performance on Benchmark Datasets

The application of metaheuristic algorithms and machine learning models is among the most effective solutions for addressing the high dimensionality of data and selecting the optimal subset of features. With the rapid growth of data generation and the diverse tools and requirements across various domains—including technology, medicine, and business—this approach has attracted considerable attention for optimizing processes and outcomes such as reducing costs and improving prediction accuracy. The Grey Wolf Optimizer (GWO), inspired by the hierarchical and social behavior of grey wolves, and the Cooperation Search Algorithm (CSA), inspired by the collaborative behavior of individuals across various organizational ranks in a modern organization, are two examples of such metaheuristic algorithms. In this study, we propose a novel approach for solving the feature selection problem using a hybrid optimization algorithm that combines GWO in the exploration phase and CSA in the exploitation phase. We evaluated the performance of our proposed hybrid model in terms of accuracy, precision, F-score, and recall using three datasets from the UCI (University of California, Irvine) Machine Learning Repository, comparing it against three baseline algorithms: CSA, GWO, and Ant Colony Optimization (ACO). The results demonstrate the superiority of the proposed hybrid model over its competitors, achieving improved performance across all evaluation metrics.

کلیدواژه ها:

Grey Wolf Optimizer ، Cooperation Search Algorithm ، Feature Selection ، Optimization ، Metaheuristic