An application of credit score modelling

Abstract

This chapter uses three standard statistical techniques, namely, discriminant analysis, logistic regression and survival analysis, to construct and evaluate a credit-scoring model for the Al-Wahada Bank in Libya. We randomly select 400 hundred loans for the analysis and specify 21 independent variables along with measures indicating loan default. The results indicate that five independent variables are common to all three models, which we consider the most important variables affecting the probability of loan default. With regard to a comparison of model performance, the results indicate that logistic regression performs better than either discriminant or survival analysis in this particular context.