A Comparative Study in Binary Classification for Loan Eligibility Prediction

Abstract

In this thesis, loan eligibility prediction is explored and analyzed by investigating five different prediction models. The goal for the prediction models is to accurately predict whether a bank loan is approved or disapproved. The dataset utilized for this thesis is retrieved from kaggel, and is referred to as “bank loan data”. It contains about 100.000 rows of various loan applications, with the predictor variable available. Throughout this thesis, we will investigate five supervised learning algorithms, more specifically, Support Vector Machine, K-Nearest Neighbors, Decision Tree, Logistic Regression, and Stochastic Gradient Descent. The results are evaluated using various performance measures, and is compared to similar research within the same topic.