Abstract: To improve the customer service level of enterprises, five customer churn prediction algorithms were summarized and three algorithm evaluation models were introduced. Customer churn reasons were concluded, and customer retention was classified. The research achievements were demonstrated in telecommunications, finance an E-commerce industry. The existing problems for current research such as less researches on integration customer information acquisition, prediction, analysis reasons, assessment and retention, less samples for generally accepted prediction, less customer churn analyses based on system dynamics, less evaluations for dynamic customer churn, and less researches for retention based on quantitative analysis were summarized. These efforts would promote the researches of customer churn theory, approach and application.

Key words: customer relationship management, customer churn, churn estimation, churn retention, prediction models

摘要： 为了提高企业的客户服务水平，总结了五种客户流失预测算法，介绍了三种算法评价模型；归纳了造成客户流失的原因；概括了客户流失挽留的分类。列举了客户流失在电信、金融和电子商务行业的应用案例。总结了当前存在的问题：整合客户信息获取、预测、流失原因分析与评估、流失挽留的研究较少；公认的预测样本相对较少；基于系统动力学的客户流失原因分析和动态客户流失原因评估不多；基于定量分析的客户流失挽留研究较少。为深入研究客户流失理论、方法和应用提供了参考。

关键词: 客户关系管理, 客户流失, 流失评估, 流失挽留, 预测模型