Category-based gene feature selection algorithm

Abstract

In gene data analysis, feature selection can retain original information, reduce redundancy, remove irrelevant features, and select the feature genes that are beneficial for classification. In order to select better minority genes, a category-based feature selection algorithm is proposed, which mixes univariate and multivariate methods. First, the Relief-F algorithm is used to calculate the score of each feature, and the genes with high scores are retained, and in order to remove redundancy, the gene categories with low classification accuracy obtained under the support vector machine classifier are removed by combining K-means algorithm, and then by Spearman rank correlation coefficient set conditions and thresholds to select genes with obvious differences between categories. Finally, support vector machine, k-nearest neighbor and decision tree classifier are used to test feature subsets. The experiment show that the algorithm can select better genes, achieve gene data dimensionality reduction and improve classification accuracy.