基于互信息的特征子集选择

Feature subset selection based on mutual information

通过互信息的思想提出一个新的评价函数来评价属性之间的相关性,并结合LV算法进行特征子集选择.结果表明,该方法对分类问题效果明显.

Abstract：A new evaluate function was proposed based on the idea of mutual information with the purpose of evaluating the relativity of attributes. It was also integrated with LV algorithm to select the best feature subset. The result indicates that the method is very effective to solve the classification problems.