Abstract: The analysis system of subscriber churn in wireless carriers was constructed based on data mining. It used Bayesian network to build the subscriber churn model from existing business data. So the market people could make corresponding policy. Experiments prove that the priority knowledge can be corrected continually by inference information of Bayesian network, and help to solve the problem of subscriber churn.

Key words: causality, Bayesian network, machine learn, subscriber churn

摘要： 移动客户流失分析系统在数据挖掘的基础上,实现了客户流失模型的管理应用。其中关键的环节是根据先验知识的因果推理和基于贝叶斯网络的因果推理进行流失客户的分析,挖掘导致流失的因素,从而辅助市场经营人员制订相应的策略。实验说明,基于贝叶斯网络推理的知识可以不断修正先验知识,获得对客户流失等问题的正确认识。

关键词: 因果推理, 贝叶斯网络, 机器学习, 客户流失