Supervised Feature Selection Based on Improved Linear Discriminant Analysis

Feature selection method advances the performance of classification models by eliminating irrelevant and redundant features. LDA(Linear Discriminant Analysis) is a classic and efficient method for feature selection by clustering. However, the conventional LDA model is difficult to solve the problem of outlier sensitivity and sparse subspace. To solve these problems, we propose a supervised feature selection method based on improved LDA model. Firstly, the fixed norm used to judge the distance between samples in LDA is replaced by a novel adaptive norm to reduce the influence of outliers. Secondly, the sparse subspace learning model is embedded into LDA to enhance the information density of selected features. Finally, an efficient iterative algorithm is proposed to solve the optimization problem. Outlier and classification experiments on 9 real-word datasets show that the proposed algorithm is superior to other algorithms in classification performance and robustness.