基于CART和自适应Boosting算法的移动通信企业客户流失预测模型

中国管理科学 ›› 2014, Vol. 22 ›› Issue (10): 90-96.

基于CART和自适应Boosting算法的移动通信企业客户流失预测模型

张玮¹, 杨善林¹, 刘婷婷²

1. 合肥工业大学管理学院, 安徽合肥 230009;
2. 富士康科技集团, 广东深圳 518109

收稿日期:2012-03-15 修回日期:2013-04-09 出版日期:2014-10-20 发布日期:2014-10-20
作者简介:张玮（1983-），男（汉族），浙江人，合肥工业大学管理学院博士研究生，研究方向：数据挖掘、决策理论.
基金资助:
国家自然科学基金重点资助项目（71331002）

Customer Churn Prediction in Mobile Communication Enterprises based on CART and Boosting Algorithm

ZHANG Wei¹, YANG Shan-lin¹, LIU Ting-ting²

1. Management School, Hefei University of Technology, Hefei 230009, China;
2. Foxconn Technology Group, Shenzhen, 518109, China

Received:2012-03-15 Revised:2013-04-09 Online:2014-10-20 Published:2014-10-20

摘要/Abstract

摘要： 客户流失问题一直以来都受到企业的重视，如何有效预测流失客户是一个重要课题。本文通过对某通信企业原始数据进行严格的数据预处理，以及利用直方图检验和卡方检验相结合的方法对模型变量进行筛选，同时采用抽样的方法选取出模型的训练样本和测试样本，并利用分类回归树算法和自适应Boosting算法生成相应的强分类器模型，仿真实验结果表明本文使用的模型在预测该通信企业的离网客户中具有较高的准确性，从模型的ROC曲线可知，该模型是一个比较理想的分类模型。另外，本文通过与其他两个模型的预测结果进行比较发现本文的集成模型具有更好的预测性能。

关键词: 客户流失, 自适应Boosting算法, CART算法, 预测

Abstract: Customer churn problems have been taking seriously by Enterprises, and how to predict churning customers effectively has been becoming an important subject. Firstly the original data of a mobile communication enterprise is preprocessed strictly, and the histogram test and the chi-square test are employed for choosing variables for the prediction model. Then a sampling method is applied to extract data for training and testing, and a strong classifier model based on Classification and Regression Tree and adaptive boosting algorithm is constructed by using training samples. At last, a simulation experiment is adopted and the results of the experiment show that the integrated model used in this paper had high prediction precision. The ROC curve presented in the paper also illustrates the model is an ideal classification model. Meanwhile, the model has been proved to have better prediction performance by comparison with the other two models.

Key words: customer churning, adaptive boosting algorithm, CART algorithm, prediction

中图分类号:

TP391

张玮, 杨善林, 刘婷婷. 基于CART和自适应Boosting算法的移动通信企业客户流失预测模型[J]. 中国管理科学, 2014, 22(10): 90-96.

ZHANG Wei, YANG Shan-lin, LIU Ting-ting. Customer Churn Prediction in Mobile Communication Enterprises based on CART and Boosting Algorithm[J]. Chinese Journal of Management Science, 2014, 22(10): 90-96.

参考文献

[1] Buckinx W, Van den Poel D. Customer base analysis: partial defection of behaviourally loyal clients in a non-contractual FMCG retail setting [J]. European Journal of Operational Research, 2005, 164(1): 252-268.

[2] Verbeke W, Dejaeger K, Martens D, et al. New insights into churn prediction in the telecommunication sector: A profit driven data mining approach [J]. European Journal of Operational Research, 2012, 218(1): 211-229.

[3] Glady N, Baesens B, Croux C. Modeling churn using customer lifetime value [J]. European Journal of Operational Research, 2009, 197(1): 402-411.

[4] Tsai C, Lu Y. Customer churn prediction by hybrid neural networks [J]. Expert Systems with Applications, 2009, 36(10): 12547-12553.

[5] 朱帮助,张秋菊,邹昊飞,等. 基于OSA算法和GMDH网络集成的电子商务客户流失预测 [J]. 中国管理科学,2011,19(5):64-70.

[6] 朱帮助. 基于SMC-RS-LSSVM的电子商务客户流失预测模型 [J]. 系统工程理论与实践,2010,30(11):1960-1967.

[7] Kisioglu P, Topcu Y. Applying Bayesian belief network approach to customer churn analysis: A case study on the telecom industry of Turkey [J]. Expert Systems with Applications, 2011, 38: 7151-7157.

[8] 罗彬,邵培基,罗尽尧, 等. 基于蚁群算法的成本敏感线性集成多分类器的客户流失研究 [J]. 中国管理科学,2010,18(3):58-67.

[9] 罗彬,邵培基,罗尽尧, 等. 基于多分类器动态集成的电信客户流失预测 [J]. 系统工程学报,2010,25(5):703-711.

[10] Chen Zhenyu, Fan Zhiping, Sun Minghe. A hierarchical multiple kernel support vector machine for customer churn prediction using longitudinal behavioral data [J]. European Journal of Operational Research, 2012, 223(2): 461-472.

[11] 夏国恩,金炜东. 基于支持向量机的客户流失预测模型 [J]. 系统工程理论与实践,2008,1(1):71-77.

[12] Huang Bingquan, Kechadi T, Buckley B, et al. A new feature set with new window techniques for customer churn prediction in land-line telecommunications [J]. Expert Systems with Applications, 2010, 37(5): 3657-3665.

[13] 应维云,覃正,赵宇, 等. SVM方法及其在客户流失预测中的应用研究 [J]. 系统工程理论与实践,2007, 27(7):105-110.

[14] 蒙肖莲,蔡淑琴,杜宽旗,等. 行业银行客户流失预测模型研究 [J]. 系统工程,2004,22(12):67-71.

[15] 罗彬,邵培基,罗尽尧, 等. 基于竞争对手反击的电信客户流失挽留研究 [J]. 管理科学学报,2011,14(8):17-33.

[16] 胡理增,陈建军. 无约束条件下多客户流失挽救最优化决策 [J]. 中国管理科学,2009,17(6):39-43.

[17] 胡理增,于信阳,张长赋,等. 基于经费约束和广义客户终身价值最大化的多客户流失挽救模型 [J]. 系统工程理论与实践,2009,29(2):63-69.

[18] Mozer M, Wolniewicz R, Grimes D, et al. Predicting subscriber dissatisfaction and improving retention in the wireless telecommunications industry [J]. IEEE Transactions on Neural Networks, 2000, 11(3): 690-696.

[19] Hwang H, Jung T, Suh E. An LTV model and customer segmentation based on customer value: A case study on the wireless telecommunication industry [J]. Expert Systems with Applications, 2004, 34(1): 231-241.

[20] Marsland S. Machine Learning [M]. Florida: Chapman & Hall/CRC, 2009.

[21] Valiant L G. A theory of the learnable [J]. Communications of the ACM, 1984, 27(11): 1134-1142.

[22] Schapire R. The strength of weak learnability [J]. Machine Learning, 1990, 5(2): 197-227.

[23] Freund Y. Boosting a weak learning algorithm by majority [J]. Information and Computation, 1995, 121(2): 256-285.

[24] Freund Y, Schapire R. A decision-theoretic generalization of on-line learning and an application to boosting [J]. Journal of Computer and System Sciences, 1997, 55(1): 119-139.

[25] Freund Y, Schapire R. Experiments with a new boosting algorithm [J]. Machine Learning, 1996, 96:148-156.

[26] Schapire R, Singer Y. Improved boosting algorithms using confidence-rated predictions [J]. Machine Learning, 1999,37(3):297-336.