Modeling Customer’s Credit Worthiness using Machine Learning Models: A Review

The risk management for financial organizations especially banks is associated with the financial distress caused due to credit worthiness of customer. The methodology to construct models for credit scoring varies from organization to organization. The evaluation and prediction of customer's credit worthiness is a key for preventing losses in banking sector. The purpose of calculating credit scoring is to categorize the applicants into good and bad application in terms of credit worthiness. To find out such behavior can be treated as a sort of Machine Learning (ML) problem. Now, the use of ML algorithms proved its significance in solving problems of various fields including credit risk detection and prediction. At the same time, the usages of ensemble classifiers in ML showcase an imperative role in building predictive models. Several researches confirm that use of ensemble classifiers show a considerable improvement in performance of various classification techniques. In this paper, we present the survey of works undertaken by several authors for determining and predicting credit risk. The survey is presented on basis of two paradigms of ML: 1. Machine Learning approach 2. Ensemble approach. We raised some relevant research questions to be addressed. The work also aims at providing future directions for development of feasible techniques for building predictive models.