Predicting Bank Loan Defaults

Abstract

In our everyday lives, loan lending plays an
important role. It powerfully promotes the economy and
the growth of consumption. A loan carries risk, is
inevitable, and may result in a financial crisis. As a
result, determining whether a person is eligible for the
loan is critical. In this research, we use the XGBoost
and Random Forest techniques to train the prediction
model and compare their accuracy. We use the
variance inflation factor and variance threshold
approaches in the feature engineering section to filter
out unnecessary features before feeding them into
XGBoost and Random Forest. In loan default
scenarios, there is no difference in prediction accuracy
between XGBoost and Random Forest because both
attain a high accuracy of roughly 0.9