Nowadays customer attrition is increasingly serious in commercial banks, particularly, high-valued customers in retail banking. Hence, it is encouraged to develop a prediction mechanism and identify such customers who might be at risk of attrition. This prediction mechanism can be considered to be a classifier. In particular, the problem of predicting risk of customer attrition can be prototyped as a binary classification task in data mining. In previous studies, a number of techniques have been introduced in (binary) classification study, i.e. artificial-based model, Bayesian-based model, case-based model, tree-based model, regression-based model, rule-based model, etc. With regards to a particular application predicting customer attrition risk for retail banking, this paper presents four principles in (classification) model selection. To support this model selection study, a set of experiments were run, based on a collection of real customer data in retail banking. These results and consequent recommendations are given in this paper.
|Cite as: Li, F., Lei, J., Tian Y., Punyapatthanakul, S. and Wang Y. J. (2011). Model Selection Strategy for Customer Attrition Risk Prediction in Retail Banking. In Proc. Australasian Data Mining Conference (AusDM 11) Ballarat, Australia. CRPIT, 121. Vamplew, P., Stranieri, A., Ong, K.-L., Christen, P. and Kennedy, P. J. Eds., ACS. 119-124 |
(local if available)