LOAN APPROVAL PREDICTION BASED ON MACHINE LEARNING APPROACH

Abstract

Banks have various goods to sell in the banking system. The major source of income and profit, however, is their credit lines. As a result, they can profit from the interest on the loans they credit. The profit or loss of a bank is mostly determined by loans, that is, whether consumers repay the loan or default. The bank can lower its Non-Performing Assets by forecasting loan defaulters. Previous research in this age has revealed that there are numerous techniques for studying the subject of loan default control. However, because accurate forecasts are critical for profit maximization, it is critical to investigate the nature of the various methodologies and compare them. In this research, the datasets used were gathered from Kaggle for training and testing. The results gotten from both datasets were compared to ascertain which algorithm could best be used for predicting loan approval and also to determine which features are most important in predicting loan approval. The different metrics of performance that were used to define the results are: Accuracy, Precision, Recall and F1-Score. Eight different algorithms were used to train the models, these are: the Logistic Regression algorithm, Random forest, Decision trees, Linear Regression, Support Vector Machine (SVM), Naïve Bayes, K-means and K Nearest Neighbors (KNN) algorithms. The final results revealed that the models generated varied outcomes. From the results shown across both datasets, Logistic regression had - 83.24% and 78.13% of  accuracy, followed by Naïve Bayes with 82.16% and 77.34% accuracy level, Random Forest