Evaluating loans using variable benchmark data envelopment analysis

Abstract

Loan officers use many business intelligence methods to screen consumer loan applications besides intuitive judgement and experience. They also use mathematical techniques such as credit-scoring models, traditional statistical models, and artificial intelligence methods such as expert systems, artificial neural systems, and fuzzy logic. This study illustrates the development of a decision support system using variable benchmark data envelopment analysis model to predicting bad loans. Further, the study also compares the performance of the DEA model with linear discriminant analysis model. The study illustrates the viability of the variable benchmark DEA model that outperforms the linear discriminant analysis model.