基于互信息和关系积理论的特征选择方法

研究互信息理论，针对其不足引进粗糙集并给出一个基于关系积理论的属性约简算法，以此为基础提出一个适用于海量文本数据集的特征选择方法。该方法使用互信息进行特征初选，利用所给的属性约简算法消除冗余，从而获得具有代表性的特征子集。实验结果表明，该特征选择方法效果良好。

Feature Selection Method Based on Mutual Information and Attribute Union Theory

This paper analyzes Mutual Information(MI) theory. According to deficiency of MI, rough set is introduced and an attribute reduction algorithm based on attribute union theory is proposed. A feature selection method based on MI and attribute union theory is presented which is suitable for massive text data sets. The method uses MI to select features, and employs the proposed attribute reduction algorithm to eliminate redundancy, so the feature subsets which are more representative can be acquired. Experimental results show that the method is effective.