摘要
为解决属性约简的诸多问题,比如基于信息熵的属性约简方法偏向多值属性的缺点,而基于属性相似度的属性约简方法偏向少值属性的不足,现提出一种将信息熵和属性相似度结合起来的新的启发式属性约简方法。实验结果表明,避免了上述两种属性约简算法的偏向性。
There are many problems of attribute reduction, for examples, information entropy-based attribute reduction method has the shortcomings of tending multi-valued attribute , on the contrary, similarity-based attribute reduction method swerves to the few value attribute. In order to solve such many problems of attribute reduction, a new heuristic attribute reduction method which unifies the information entropy and the attribute similarity is presented. The experiment result indicates that the new algorithm avoids the deviation of the two attribute reduction algorithms above.
出处
《科学技术与工程》
2009年第16期4809-4810,4819,共3页
Science Technology and Engineering
关键词
属性约简
信息熵
属性相似度
attribute reduction information entropy attribute similarity
