Logistic Regression or Neural Network? Which Provides Better Results for Retail Loans?

Abstract

While there is extensive literature on the prediction of corporate bankruptcies, there is little literature on the classification of retail borrowers. This is also true in Hungary. Recognising who is at risk of becoming a bad debtor is not easy. There are several ways to analyse the data, which may yield different results. In this paper, my aim is to predict the default of household loans using logistic regression and neural networks. The question is, which method produces the better results?The analyses show that the neural network model produced the best and most favourable results. The accuracy of the best method was found to be 81.5%.