非合约型客户终身价值的稳健性度量：经典方法与机器学习算法的综合测算研究

›› 2019, Vol. 31 ›› Issue (4): 83-98.

非合约型客户终身价值的稳健性度量：经典方法与机器学习算法的综合测算研究

成栋¹, 孙莹璐¹, 薛薇^2,3

1. 中国人民大学商学院, 北京 100872;
2. 中国人民大学应用统计科学研究中心, 北京 100872;
3. 中国人民大学统计学院, 北京 100872

收稿日期:2018-07-25 出版日期:2019-04-28 发布日期:2019-04-26
通讯作者: 薛薇,中国人民大学应用统计科学研究中心、中国人民大学统计学院副教授,硕士生导师,博士。
作者简介:成栋,中国人民大学商学院教授,博士生导师,博士;孙莹璐,中国人民大学商学院博士研究生
基金资助:
中国人民大学"中央高校建设世界一流大学（学科）和特色发展引导专项资金"。

Robust CLV Measurement in Non-contractual Settings:A Study of CLV Measurement Combining Probability Models and Machine Learning Algorithms

Cheng Dong¹, Sun Yinglu¹, Xue Wei^2,3

1. Business School, Renmin University of China, Beijing 100872;
2. Center for Applied Statistics of Renmin University of China, Beijing 100872;
3. School of Statistics, Renmin University of China, Beijing 100872

Received:2018-07-25 Online:2019-04-28 Published:2019-04-26

摘要/Abstract

摘要：

客户终身价值（CLV）是企业进行客户关系管理的基础，然而非合约关系下客户终身价值的度量一直是研究的难点。本文重点探讨了以Pareto/NBD和BG/NBD为代表的经典概率模型和以GAM和SVM为代表的机器学习算法在非合约客户终身价值度量中的应用。通过对两个数据集的实证研究，对比了四种方法的特点和预测能力。研究发现经典概率模型的预测值较为平稳，适用于描述消费者日常消费规律；GAM则对数据中极端变化的捕捉跟踪能力较强，适用于预测由于门店促销、线上促销和节假日等带来的不规律的集中消费或延时消费的情况。经典方法和机器学习算法对客户终身价值的预测各有所长，基于单一方法的预测会有一定偏差，为得到小偏差和高稳健性的CLV估计，本文认为基于多方法的综合预测是理想的CLV建模策略。

关键词: 客户终身价值(CLV), Pareto/NBD, BG/NBD, 广义可加模型(GAM), 支持向量机(SVM)

Abstract:

Customer lifetime value (CLV) is a key concept in customerrelationship management. However, measuring CLV, especially CLV in non-contractual settings is one of the most challenging research topics. In this paper, we explore two probability models (Pareto/NBD and BG/NBD) and two machine learning models (GAM and SVM) in CLV modeling. Through two empirical studies, we compare the characteristics and prediction performance of the four mentioned approaches. The study shows that the prediction performance of classical probability models is very stable, which means that probability models are able to capture the regular patterns of consumers' daily consumption. While GAM has a good performance in capturing extreme data, making that an advantage to reflect irregular consumptions such as stock up resulted from store promotion, online promotion or holidays and festivals. Since each approach has its own strong points, we believe that an integrated CLV measurement combining probability models and machine learning algorithms is an ideal solution to ensure the robustness of CLV measurement.

Key words: customer lifetime value (CLV), Pareto/NBD, BG/NBD, generalized additive model (GAM), support vector machine (SVM)

成栋, 孙莹璐, 薛薇. 非合约型客户终身价值的稳健性度量：经典方法与机器学习算法的综合测算研究[J]. , 2019, 31(4): 83-98.

Cheng Dong, Sun Yinglu, Xue Wei. Robust CLV Measurement in Non-contractual Settings:A Study of CLV Measurement Combining Probability Models and Machine Learning Algorithms[J]. , 2019, 31(4): 83-98.

参考文献

[1] 刘东昌. 客户关系管理(CRM)与客户选择[J]. 管理评论, 2003,15(7):41-44
[2] Reinartz W., Kumar V. The Impact of Customer Relationship Characteristics on Profitable Lifetime Duration[J]. Journal of Marketing, 2003,67(1):77-99
[3] Payne A., Frow P. A Strategic Framework for Customer Relationship Management[J]. Journal of Marketing, 2005,69(4):167-176
[4] Venkatesan R., Kumar V. A Customer Lifetime Value Framework for Customer Selection and Resource Allocation[J]. Journal of Marketing, 2004,68(4):106-125
[5] Schmittlein D. C., Morrison D. G., Colombo R. Counting Your Customers:Who Are They and What Will They Do Next[J]. Management Science, 1987,33(1):1-24
[6] ThomasJ. S., Blattberg R. C., Fox E. J. Recapturing Lost Customers[J]. Journal of Marketing Research, 2004,41(1):31-45
[7] Gupta S., Hanssens D., Hardie B., et al. Modeling Customer Lifetime Value[J]. Journal of Service Research, 2006,9(2):139-155
[8] Estrella-Ramón A. M., Sanchez-Perez M., Swinnen G., et al. A Marketing View of the Customer Value:Customer Lifetime Value and Customer Equity[J]. South African Journal of Business Management, 2013,44(4):47-64
[9] Coussement K., Benoit D. F., Van Den Poel D. Improved Marketing Decision Making in a Customer Churn Prediction Context Using Generalized Additive Models[J]. Expert Systems with Applications, 2010,37(3):2132-2143
[10] 应维云. 随机森林方法及其在客户流失预测中的应用研究[J]. 管理评论, 2012,24(2):140-145
[11] Reinartz W., Kumar V. On the Profitability of Long-life Customers in a Non-contractual Setting:An Empirical Investigation and Implications for Marketing[J]. Journal of Marketing, 2000,64(4):17-35
[12] Kumar V., Reinartz W. Creating Enduring Customer Value[J]. Journal of Marketing, 2016,80(6):36-68
[13] Berger P. D., Nasr N. I. Customer Lifetime Value:Marketing Models and Applications[J]. Journal of Interactive Marketing, 1998,12(1):17-30
[14] Bult J. R., Wansbeek T. Optimal Selection for Direct Mail[J]. Marketing Science, 1995,14(4):378-394
[15] Fader P. S., Hardie B. G. S., Lee K. L. RFM and CLV:Using Iso-value Curves for Customer Base Analysis[J]. Journal of Marketing Research, 2005,42(4):415-430
[16] Dwye F. R. Customer Lifetime Valuation to Support Marketing Decision Making[J]. Journal of Direct Marketing, 1997,11(4):7-13
[17] Fader P. S., Hardie B. G. S., Lee K. L. Counting Your Customers the Easy Way:An Alternative to the Pareto/NBD Model[J]. Marketing Science, 2005,24(2):275-284
[18] Schmittlein D. C., Peterson R. A. Customer Base Analysis:An Industrial Purchase Process Application[J]. Marketing Science, 1994,13(1):41-67
[19] Colombo R., Jiang W. A Stochastic RFM Model[J]. Journal of Interactive Marketing, 1999,13(3):2-12
[20] Boatwright P., Borle S., Kadane J. B. A Model of the Joint Distribution of Purchase Quantity and Timing[J]. Journal of the American Statistical Association, 2003,98(463):564-572
[21] Venkatesan R., Kumar V., Bohling T. Optimal Customer Relationship Management Using Bayesian Decision Theory:An Application for Customer Selection[J]. Journal of Marketing Research, 2007,44(4):579-594
[22] Borle S., Singh S. S., Jain D. C. Customer Lifetime Value Measurement[J]. Management Science, 2008,54(1):100-112
[23] Erevelles S., Fukawa N., Swayne L. Big Data Consumer Analytics and the Transformation of Marketing[J]. Journal of Business Research, 2016,69(2):897-904
[24] Lemmens A., Croux C. Bagging and Boosting Classification Trees to Predict Churn[J]. Journal of Marketing Research, 2006,43(2):276-286
[25] Coussement K., Van Den Poel D. Churn Prediction in Subscription Services:An Application of Support Vector Machines While Comparing Two Parameter-selection Techniques[J]. Expert Systems with Applications, 2008,34(1):313-327
[26] Chen Z. Y., Fan Z. P. Dynamic Customer Lifetime Value Prediction Using Longitudinal Data:An Improved Multiple Kernel SVR Approach[J]. Knowledge-based System, 2013,43(5):123-134
[27] Giuffrida G., Chu W. W., Hanssens D. M. Mining Classification Rules from Datasets with Large Number of Many-valued Attributes[DB/OL]. https://doi.org/10.1007/3-540-46439-5_23, 2000
[28] 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, 2008,28(1):71-77
[29] Vafeiadis T., Diamantaras K. I., Sarigiannidis G., et al. A Comparison of Machine Learning Techniques for Customer Churn Prediction[J]. Simulation Modelling Practice and Theory, 2015,55(6):1-9
[30] Martínez A., Schmuck C., Pereverzyev Jr. S., et al. A Machine Learning Framework for Customer Purchase Prediction in the Non-contractual Setting[J/OL]. https://doi.org/10.1016/j.ejor.2018.04.034, 2018
[31] Chen P., Guitart A., Fernández del Río A., et al. Customer Lifetime Value in Video Games Using Deep Learning and Parametric Models[DB/OL]. https://arxiv.org/pdf/1811.12799.pdf, 2018
[32] Sifa R., Runge J., Bauckhage C., et al. Customer Lifetime Value Prediction in Non-Contractual Freemium Settings:Chasing High-Value Users Using Deep Neural Networks and SMOTE[C]. The 51st Hawaii International Conference on System Science, 2018
[33] Coussement K., Lessmann S., Verstraeten G. A Comparative Analysis of Data Preparation Algorithms for Customer Churn Prediction:A Case Study in the Telecommunication Industry[J]. Decision Support Systems,2017,95(3):27-36
[34] Hastie T., Tibshirani R. Generalized Additive Models[J]. Statistical Science, 1986,1(3):297-318
[35] Vapnik V. Statistical Learning Theory[M]. New York:Wiley Press, 1998