关键词: 数据挖掘, 贝叶斯决策树, 客户流失, 熵函数

Abstract: According to telecom enterprise customer churn problem, a prediction model based on Bayesian decision tree algorithm is put forward. It combines the prior information method of Bayesian classification and the information gain method of decision tree classification, and applies to the analysis of telecom enterprise customer churn, then puts the customer data of mobile company and data UCI in the model respectively and gets the relevant results. Added Bayesian node to make up for the decision tree cannot handle the missing value and the ambiguity data. The applications indicate that, Bayesian decision tree algorithm sacrifices a little training time and classification time, but gets higher coverage rate and hit rate than basic decision tree method.

Key words: data mining, Bayesian decision tree, customers churn, entropy function