Feature Selection Algorithm Based on Maximum Information Coefficient

Abstract

Feature selection is the key technology to improve the performance of classification learning problem with a large feature size. In this paper, maximum information coefficient (MIC) is introduced to measure the correlation between each feature and the decision. We design a feature selection algorithm based on combining maximum information coefficient (MIC) and classification learning model. Firstly, the maximum information coefficient (MIC) is calculated to get the correlation degree between each feature and the decision. Furthermore, according to the importance, the feature ranking results are obtained. Secondly, combining with the specific classification learning model, a feature selection method based on the maximum information coefficient (MIC) is constructed. Taking the classification accuracy as the estimation basis, the optimal feature subset is selected. Then, we employ the proposed algorithm to a real-world classification problem. Experimental results show that the feature subset generated by the proposed method can effectively select sensitive features, reduce the dimension of feature space, and achieve higher classification accuracy.