Feature Selection Method Based on Improved Differential Evolution and ReliefF
Manting Yan, Jinyu Deng, Shuye Zhang, Pengyu Chen · 2024
In addressing the challenge of difficult coupled feature selection in high-dimensional sample environments, this paper proposes a feature selection method based on a combination of improved differential evolution and the ReliefF algorithm (MRDE). Initially, the ReliefF algorithm is employed to filter out features with weaker impacts on classification results. Subsequently, the remaining features undergo feature extraction using the improved differential evolution algorithm. The enhanced diversity of feature set selection is achieved through the improved cross-strategy of the differential evolution algorithm, thereby resolving the issue of multi-coupled feature selection. Experimental results demonstrate the effectiveness of the proposed algorithm in feature extraction, with extracted features exhibiting strong classification accuracy.