An Integrated Genetic-based Model of Naive Bayes Networks for Credit Scoring

Inappropriate management in some fields such as credit allocation has imposed too many losses to financial institutions and even has forced some of them to go bankrupt. Moreover, large volume data sets collected by credit departments has necessitated utilizing highly accurate models with less complexities. Credit scoring models with classification and forecasting customers into two groups good and bad can dramatically reduce risks of granting credits to customers. In this paper, a novel integrated approach for credit scoring problem is presented. This approach utilizes rough sets for feature selection during the data pre-processing phase and also adopts two hybrid sequences, Naïve Bayes networks and genetic algorithm, to classify customers. In order to assess the competitive performance of the proposed approach, it has been executed on three credit scoring datasets from the University of California Irvine Machine Learning Repository. Computational results demonstrate that our approach has superior performance in terms of classification accuracy and achieves higher overall classification rate as compared to several other previous studies.