Elite-driven grey wolf optimization for global optimization and its application to feature selection

Feature selection is crucial in data preprocessing, especially in medical data analysis. Although the Grey Wolf Optimization (GWO) algorithm has attracted attention because of its simplicity and efficiency, it is prone to falling into the local optimum when searching fora globally optimal solution when dealing with complex feature selection problems, which restricts its application potential. To solve this problem, this paper proposes the Elite-driven Grey Wolf Optimizer (EDGWO) algorithm. The EDGWO algorithm significantly improves the global search capability of Alpha, Beta, and Delta grey wolves by taking advantage of the social hierarchy of the grey wolf population and designing three global exploration operators. The algorithm smoothly transitions from extensive exploration to intensive exploitation by dynamically adjusting the search parameters A. In addition, the introduced stochastic probabilistic search strategy allows omega grey wolves to make a flexible choice between local exploitation and global exploration, effectively avoiding premature convergence during the search process. To evaluate the performance of the EDGWO algorithm, this study compared twenty-two standard benchmark functions of CEC2021 and CEC2022 and twelve cancer microarray datasets. The experimental results show that the EDGWO algorithm demonstrates superior exploration and exploitation capabilities compared to fifteen well-known algorithms, with fast convergence speed and effective circumvention of local optima. Various evaluations have shown that EDGWO achieved the best Friedman rankings in the 10- and 20-dimensional CEC2021 and CEC2022 benchmark functions. In particular, the EDGWO algorithm maintains high convergence speed and high accuracy in feature selection for cancer microarray datasets.