Loan Default Prediction using Machine learning Techniques

Abstract

Loans are a very fundamental source of any bank’s revenue, so they work tirelessly to make sure that they only give loans to customers who will not default on the monthly payments. They pay a lot of attention to this issue and use various ways to detect and predict the default behaviours of their customers. However, a lot of the time, because of human error they may fail to see some key information. The main objective of this work is to automate the process of loan default prediction by using machine learning algorithms like K-Nearest Neighbours, Decision Tree, Support Vector Machine and Logistic Regression to predict defaulters. The accuracy of these methods will also be tested using metrics like Log Loss, Jaccard Similarity Coefficient and F₁ Score. These metrics are compared to determine the accuracy of prediction. This can help banks conserve their manpower and fiscal resources by reducing the number of steps they have to take in order to check if somebody is eligible for a loan.