Using adaptive learning in credit scoring to estimate acceptance probability distribution

Credit Scoring and Behaviour Scoring are tools that are widely used in the applications of quantitative analysis in businesses. The main purpose is to assess the risk of such customers defaulting but in the commercial environment. Currently in consumer lending, there is an increasing need to assess whether the customer is most likely to accept a variant of a product. Thus, if one wishes to use the scorecard both for risk assessment and for product acceptance, one may wish to vary the product offered and the questions asked in order to improve the estimates of the probability of acceptance as a function of the features offered. In the beginning phase, we will look at the strategies that only change one feature of the product offered, so as to improve the overall profitability of the product. This includes assessing and updating the probability of the customer accepting the various versions of the product. The underlying model that was built was an adaptive dynamic programming model with elements of Bayes theorem to include previous actions. The subsequent work was to look at the possibilities of adjusting the questions asked in the scorecard to improve the estimates of customer acceptance without diminishing the credit risk assessment. Applicants for credit have to provide information for the risk assessment process. In the current conditions of a saturated consumer lending market, and hence falling take rates, information like this can be used to assess the probability of a customer accepting the offer. Also, lenders do not want to make the application process too complicated, and with the growth in adaptive marketing channels like the Internet and the telephone, they can make the questions they ask depend on the previous answers. We investigated how one could develop such ‘adaptive’ application forms; which would assess acceptance probabilities as well as risk of default. Finally, we look at a model which looks both at which question to ask and what offer to make when the answer to the question is known. This extends the work of the earlier dynamic programming model.