A Binary Multi-objective Grey Wolf Optimization for Feature Selection

In this paper, a multi-objective feature selection method based on Grey Wolf Optimization (GWO), named BMOGWO-FS, is proposed. Specifically, this paper first introduces a binary multi-objective GWO (BMOGWO), considering that feature selection problem is a 0–1 integer programming. Then, on the basis of BMOGWO, a multi-objective feature selection method named BMOGWO-FS is proposed, aiming to minimize the number of selected features while maximizing classification accuracy. To validate the performance of BMOGWO-FS, six different classifiers are employed, and a comparative analysis is conducted against six existing heuristic algorithms. Experimental results demonstrate that BMOGWO-FS achieves the best performance while maintaining good robustness and stability. Furthermore, BMOGWO-FS is applied to a real-world dataset of endometrial cancer. The experimental results show a significant improvement in the accuracy of predicting endometrial cancer recurrence after employing the BMOGWO-FS algorithm for feature selection.