AN ANALYSIS OF CREDIT DECISIONS FOR CONSUMER LOANS ON AN ONLINE PLATFORM USING TRADITIONAL AND MACHINE LEARNING TECHNIQUES

Abstract

This paper is aimed to analyse the secondary bank credit scoring mechanism using the historical data of the existing loan applications of the borrowers through traditional logistic regression and machine learning techniques. The historical data include the four-month long loan application by the clients and consists of the data of borrower’s age, sex, region, mobile model, loan overdue information, whether the borrower has a tax debt and whether the bank has refused to provide a loan to a specific client or not. In this paper, the different methodologies and the importance of machine learning in credit scoring nowadays is discussed first. Then, the description of the data and the methodology used is presented. The results include the logistic regression analysis of the variables and the check on the significance of the variable in decision-making in credit scoring. Lastly, using logistic regression and machine learning (xgboost), I was able to identify which of the strategies are better in determining the area under the curve.