An Improved Maximum Relevance and Minimum Redundancy Feature Selection Algorithm Based on Normalized Mutual Information

摘要

We present in this paper a comprehensive analysis of the mutual information based feature selection algorithms. We point out the limitations of some recent work in this area then propose an improvement to overcome the weak points. The experiment results confirm that we achieve a better feature sets compared with the two recent developed algorithms, which are Maximum Relevance and Minimum Redundancy (mRMR) and Normalized Mutual Information Feature Selection (NMIFS), in terms of the classification accuracy.