Loan Eligibility Prediction Using Ensemble Machine Learning Techniques and SMOTE

Loans play important roles in banking sectors. Loans provide bank income and fuel growth for the banking industry itself. Loans that are provided to the lender should be repaid on schedule to the bank. Therefore, banks need to carefully evaluate each customer whether they will repay the loan back. However, some institutions find difficulties to accurately predicting whether the customers are good defaulters or bad defaulters. With the existence of machine learning technologies, many financial institutions seek cutting-edge technologies. To simplify the process of approving customers’ eligibility for the loan banking system. In this paper, we will use Logistic Regression (LR), K-Nearest Neighbors (KNN), Decision Tree (DT), Random Forest (RF), XGBoost (XGB), and LightGBM (LGBM) to improve loan eligibility predictions. With this model we are using several data preprocessing methods such as handling missing values, Handling outliers, one-hot encoding, and oversampling (SMOTE). Overall evaluation metrics will be measured with Accuracy, Recall, Precision, and F1 scores to determine which model provides the highest Score. In this paper, our XGBoost model gives the best output measured with more than 95% in every evaluation metric.