Feature Selection Method Based on Chi-Square Test and Minimum Redundancy

This paper studies the feature selection problem of high-dimensional classification sample data, and proposes a feature selection method based on the combination of chi-square test and minimum redundancy. Firstly, the chi-square test is used to select the sample data highly related to the classes and reduce the data scale quickly and effectively. On this basis, the minimum redundancy algorithm is used to further remove the redundancy and realize feature selection. In order to verify the feasibility of the method proposed in this paper, the method is applied to the feature selection of ALT-ALB-AML, Breast-A and Stomach data sets. By using support vector machine, K-nearest neighbor classification and geometric barycenter classification to predict and classify high-dimensional classification sample data, it is found that the feature selection method presented in our paper can effectively classify and improve classification accuracy compared with chi-square test or minimum redundancy-maximum relevance or simple combination of the two.