ESTIMATION OF PROBABILITY OF DEFAULT OF A FINANCE HOUSE IN NIGERIA: A COMPARATIVE ANALYSIS BETWEEN LOGISTIC REGRESSION AND NEURAL NETWORK MODEL

Abstract

Effective credit management is essential for economic stability, while poor management can have destabilizing effects. Accurate credit risk assessment, including probability of default (PD), is crucial for financial institutions to prevent substantial losses from non-performing loans. The purpose of the study is to investigate the probability of default in low default portfolio of a finance house in Nigeria. In order to fulfil this purpose, two different models for estimating the probability of default, the logistic regression and the neutral network models, were considered. The data analysis was steered through SPSS version 23 software. Through a comparative analysis of logistic regression and neural network models, we determined that the neural network model is superior in predicting the probability of default. The study findings unveiled that the primary factors influencing the probability of default are the Tenor of loan and Annual nominal interest rate. The study recommends considering adjustments to terms to mitigate risk, adjusting risk strategies and implementing the neutral network model for ongoing prediction of default probabilities.