Predicting Loan Repayment: A Machine Learning Approach

Abstract

In order to reduce the risk of loan payment default, banks must assess and forecast the loaners' capacity to repay. As a result, a system has been established by the banks to handle the loan application according to the loaner's circumstances, including work status, credit history, etc. However, some loaners, including students or those without credit histories, may not be able to be evaluated for repayment capacity using the present rating approach. We trained a variety of machine learning models on the Home Credit Default Risk Kaggle dataset and assessed the significance of each feature in order to accurately determine the repayment capacity of every group of individuals. Next, using the features relevance score as a guide, we examine and choose the most recognizable characteristics to forecast the loaner's capacity for payback.