基于迁移学习的客户信用评估模型研究

doi:10.12005/orms.2015.0065

摘要/Abstract

摘要： 客户信用评估是银行等金融企业日常经营活动中的重要组成部分。一般违约样本在客户总体中只占少数,而能按时还款客户样本占多数,这就是客户信用评估中常见的类别不平衡问题。目前,用于客户信用评估的方法尚不能有效解决少数类样本稀缺带来的类别不平衡。本研究引入迁移学习技术整合系统内外部信息,以解决少数类样本稀缺带来的类别不平衡问题。为了提高对来自系统外部少数类样本信息的使用效率,构建了一种新的迁移学习模型:以基于集成技术的迁移装袋模型为基础,使用两阶段抽样和数据分组处理技术分别对其基模型生成和集成策略进行改进。运用重庆某商业银行信用卡客户数据进行的实证研究结果表明:与目前客户信用评估的常用方法相比,新模型能更好地处理绝对稀缺条件下类别不平衡对客户信用评估的影响,特别对占少数的违约客户有更好的预测精度。

关键词: 客户信用评估, 类别不平衡, 迁移学习, 数据分组处理技术

Abstract: Customer credit scoring is an important part of daily business activities for financial companies such as banks. Default customers usually makae up the minority of the population while customers of timely repayment make up the majority, which is called a class imbalance problem in the study of customer credit scoring. Existing methods in credit scoring cannot effectively solve the issue of class imbalance caused by absolute scarcity of the minority class. In our study, we introduce the technique of transfer learning to integrate the external information and try to solve the issue of class imbalance caused by absolute scarcity of the minority class. In order to exploit the minority sample outside the system more effectively, a transfer learning model is proposed, which is based on the ensemble transfer learning technology transfer bagging. A two-stage sampling method and the technique of group method of data handling are used in the new model to improve the generation and integration strategy of base models. The empirical results on the credit card dataset from a commercial bank show that the new model can deal with the issue of class imbalance caused by absolute scarcity better in comparison with other commonly used methods in credit scoring and provide a better prediction of the credit status of default customers.

Key words: credit scoring, class imbalance, transfer learning, group method of data handling

中图分类号:

C931

朱兵, 贺昌政, 李慧媛. 基于迁移学习的客户信用评估模型研究[J]. 运筹与管理, 2015, 24(2): 201-207.

ZHU Bing, HE Chang-zheng, LI Hui-yuan. Research on Credit Scoring Model Based on Transfer Learning[J]. Operations Research and Management Science, 2015, 24(2): 201-207.

参考文献

[1] Crook J N, Edelman D B, Thomas L C. Recent developments in consumer credit risk assessment[J]. European Journal of Operational Research, 2007, 183(3): 1447-1465.
[2] Orgler Y E. A credit scoring model for commercial loans[J]. Journal of Money, Credit and Banking, 1970, 2(4): 435-445.
[3] 刘京礼,李建平,徐伟宣,石勇.信用评估中的鲁棒赋权自适应Lp最小二乘支持向量机方法[J].中国管理科学,2010,5:28-33.
[4] Chang S Y, Yeh T Y. An artificial immune classifier for credit scoring analysis[J]. Applied Soft Computing, 2012, 12(2): 611-618.
[5] 肖智,李文娟.RS-ANN在消费信贷个人信用评估中的实证研究[J].软科学,2011,25(4):141-144.
[6] West D, Dellana S, Qian J. Neural network ensemble strategies for financial decision applications[J]. Computers & Operations Research, 2005, 32: 2543-2559.
[7] Haibo H, Garcia E. Learning from imbalanced data[J]. IEEE Transactions on Knowledge and Data Engineering, 2009, 21(9): 1263-1284.
[8] Sun Y, Wong A, Kamel M, Classification of imbalanced data: a review[J]. International Journal of Pattern Recognition and Artificial Intelligence, 2009, 23(4): 687-719.
[9] Chawla N V, Bowyer K W, Hall L O, Kegelmeyer W P. SMOTE: Synthetic minority over-sampling technique[J]. Journal of Artificial Intelligence Research, 2002, 16: 321-357.
[10] 吴冲,夏晗.基于支持向量机集成的电子商务环境下客户信用评估模型研究[J].中国管理科学,2008,S1.
[11] Yang Y. Adaptive credit scoring with kernel learning methods[J]. European Journal of Operational Research, 2007, 183(3): 1521-1536.
[12] Skinner B F. Science and human behavior[M]. Colliler-Macmillian, 1953.
[13] Thrun S. Is learning the N-th thing any easier than learning the first?[C]. In: Proc. of NIPS-96, 1996. 640-646.
[14] Baxter J. A Bayesian/information theoretic model of learning to learn via multiple task sampling[J]. Machine Learning, 1997, 28(1): 7-39.
[15] Caruana R. Multitask learning[J]. Machine Learning, 1997, 28 : 41-75.
[16] Ben-David S, Schuller R. Exploiting task relatedness for multiple task learning[C]. In: Proc. 16th Annual Conference on Computational Learning Theory, Washington, DC, USA, 2003.
[17] Mahmud M, Ray S R. Transfer learning using Kolmogorov complexity: basic theory and empirical evaluations[C]. In: Proc. of the 20th Annual Conference on Neural Information Processing Systems, Cambridge, MA: MIT Press, 2008. 985-992.
[18] Pan S J, Yang Q. A survey on transfer learning[J]. IEEE Transactions on Knowledge and Data Engineering, 2010, 22(10): 1345-1359.
[19] Dai W Y, Yang Q, Xue G R, Yu R. Boosting for transfer learning[C]. In: Proc. of the 24th International Conference on Machine Learning, ACM Press, 2007.193-200.
[20] Kamishima T, Hamasaki M, Akaho S. TrBagg: A simple transfer learning method and its application to personalization in collaborative tagging[C]. In: Proc. of Ninth IEEE International Conference on Data Mining, 2009. 219-228.
[21] Wang G, et al. A comparative assessment of ensemble learning for credit scoring[J]. Expert Systems with Applications, 2011, 38: 223-230.
[22] Weiss G M. Mining with rarity: a unifying framework[J]. ACM SIGKDD Explorations Newsletter, 2004, 6(1): 7-19.
[23] Mueller J A, Lemke F. Self-organizing data mining: an intelligent approach to extract knowledge from data[M]. Hamburg : Libri Books, Berlin, 2000.
[24] Xiao J, He C Z, Jiang X Y, Liu D H. A dynamic classifier ensemble selection approach for noise data[J]. Information Science, 2010, 180(18): 3402-3421.
[25] Kira K, Rendell L A. The feature selection problem: traditional methods and a new algorithm[C]. In: Proc. of Tenth National Conference on Artificial Intelligence, MIT Press, 1992. 129-134.