The Use of Machine Learning Algorithms for Bank Loan Prediction

Abstract

Background: Bank loan prediction is an important problem in the banking industry. By using historical data and applying predictive models, banks can identify patterns and make accurate predictions about loan defaults. This can help them make informed decisions about lending and minimize their losses.

Objectives: To study the important parameters that influence loan and to predict the bank loan using machine learning algorithms

Methods: The CRISP-DM process is a comprehensive and structured approach to developing predictive models. By following this process, the study can ensure that all necessary steps are taken to develop an accurate and reliable predictive model for personal loan. The use of three machine learning algorithms such as decision tree, naïve bayes, and support vector machine can provide for developing the model and enable the study to select the best one.

Results: The results suggest that the J48 Decision Tree algorithm achieved the highest accuracy of 98.85%, followed by the SVM algorithm with an accuracy of 94.01%, and the Naive Bayes algorithm with an accuracy of 89.53%. In terms of precision, recall, and F-measure, all three algorithms achieved similar performance, with values ranging from 0.895 to 0.989.

Conclusions: The performance of different machine learning algorithms in predicting bank loan showed that J48 DT was the most appropriate algorithm for developing a bank loan predictor, based on its high accuracy, low mean absolute error, and fast training time. To improve the accuracy and applicability of the model, it may be necessary to collect additional data or refine the feature selection process to identify the most relevant attributes.