Attribute reduction based on the principle of maximal dependency and minimal mutual information

Attribute reduction plays key role in the process of extracting classification rules with rough set technique. Method based on attribute significance has been extensively studied for finding a reduct. However, this method only selects the most significant attributes and do not consider the mutual relevance among the attributes in the reduct. This paper proposes a novel attribute reduction method based on the principle of maximal dependency between decision attribute and condition attributes and minimal mutual information. We conduct several experiments and compare with benchmark reduction method based on dependency. The experimental results show that our proposed method is feasible and effective. Especially, it can improve classification accuracy.