Application of classification methods in loan approval in Banking

Abstract

This research focuses on applying machine learning to predict loan approval in the financial sector. The methodology includes collecting data from the "Loan-Approval-Prediction-Dataset" on Kaggle, preprocessing the data to standardize and remove noise, and then applying classification models such as Logistic Regression, Naive Bayes, and Decision Tree using Orange software. The dataset is split into 70% for training and 30% for testing to ensure objectivity in model evaluation. Key performance metrics include Accuracy, Precision, Recall, F1-Score, and Confusion Matrix. The results show that the Decision Tree model outperforms other models, achieving an accuracy (CA) of 97.7% and an AUC of 0.976, surpassing Logistic Regression and Naive Bayes. Analysis of factors influencing loan approval reveals that credit score is the most significant factor, with scores above 549 significantly increasing approval chances. Other factors, such as commercial and residential asset values, loan term, number of dependents, and annual income, also impact approval decisions. Based on these findings, the study recommends that banks prioritize customers with high credit scores, substantial asset values, and shorter loan terms to minimize credit risk. Additionally, optimizing the approval process to reduce Type 2 errors (rejecting eligible customers) is crucial. Future research directions include expanding the dataset from multiple sources and integrating an automated prediction system into real-world banking operations