Forecasting Probability of Default for Consumer Loan Management with Gaussian Mixture Models

TL;DR

This paper introduces a Gaussian Mixture Model-based approach for predicting individual loan default probabilities, providing a probabilistic assessment that enhances decision-making and risk management in financial institutions.

Contribution

The paper presents a novel GMM-based method for probabilistic default prediction, improving interpretability and computational efficiency over traditional binary classification models.

Findings

Probabilistic default estimates closely match actual losses.

Model offers a more informative alternative to binary credit scoring.

Approach is computationally efficient for large datasets.

Abstract

Credit scoring is an essential tool used by global financial institutions and credit lenders for financial decision making. In this paper, we introduce a new method based on Gaussian Mixture Model (GMM) to forecast the probability of default for individual loan applicants. Clustering similar customers with each other, our model associates a probability of being healthy to each group. In addition, our GMM-based model probabilistically associates individual samples to clusters, and then estimates the probability of default for each individual based on how it relates to GMM clusters. We provide applications for risk managers and decision makers in banks and non-bank financial institutions to maximize profit and mitigate the expected loss by giving loans to those who have a probability of default below a decision threshold. Our model has a number of advantages. First, it gives a probabilistic view of credit standing for each individual applicant instead of a binary classification and therefore provides more information for financial decision makers. Second, the expected loss on the train set calculated by our GMM-based default probabilities is very close to the actual loss, and third, our approach is computationally efficient.