Two-Stage Decomposition Method Based on Cooperation Coevolution for Feature Selection on High-Dimensional Classification

Abstract

Feature selection is an important task in data mining and machine learning to reduce the dimensionality of data and improve the performance. However, feature selection is still a challenge task especially for the large-scale problems with small sample size and extremely large number of features. A variety of methods have been applied to solve the feature selection problems, in which evolutionary algorithm has recently attracted increasing attention and made great progress. In this study, a two-stage decomposition cooperating coevolution strategy for feature selection (CCFS/TD) is proposed. In the first stage, the proposed algorithm decomposes evolutionary process into $k$ -level and evolves by cooperating coevolution for each level. Then, in the second stage, evolution process of each level is further decomposed into several independent processes. The selected subset of features are determined by the results of all independent processes through majority voting. Experiments on ten benchmark datasets are carried out to verify the effectiveness of the proposed method. The results demonstrate that the proposed CCFS/TD can obtain better classification performance with a smaller number of features in most cases in comparison to some existing methods.