一种基于信息熵的聚类结果评价方法

A clustering results evaluation method based on information entropy

文章基于信息熵理论,将模糊聚类评价方法和决策树分类算法联系起来,提出从信息熵角度用决策树分类算法来评价聚类算法结果的有效性,从而确定最佳的聚类个数;并将该方法应用到证券行业客户忠诚度分析模型的建立中。实验结果表明,该方法可以明显提高聚类的效果,并且使得聚类结果的可解释性强,具有良好的实用价值。

In this paper,based on the theory of information entropy,the decision tree classification method is introduced to evaluate the effectiveness of the results of fuzzy clustering algorithm.In this way,the optimal number of clustering can be determined.This evolution method is applied to establishing the customer fidelity model for securities busines.The experiment results show that the method is practical,and can markedly improve the clustering efficiency and make the clustering results more interpretable.