Dynamic credit scoring on consumer behavior using fuzzy Markov model

The present financial tsunami has brought credit risk into a main focus. Dynamic credit scoring tools is highly demanded by commercial banks with products like credit cards. However, till now practitioners almost only employ statistical scoring models such as regressions. Thus the purpose of this paper is to provide for a new direction of attempts in modeling consumer credit risk and behavioral scoring dynamics. The model features heterogeneity across consumers and over time, which is realized by such a process that the credit migration rate matrix of the fuzzy Markov chain is inferred by the fuzzy inference system based on a reasonable setting of rules for each consumer, and updated at each time. A training procedure based on maximum likelihood criterion is developed for model parameters' estimation. In addition to dynamic behavioral scoring and credit behavior forecast, the model can also evaluate credit quality of each consumer according to two indicators: credit level and credit volatility.