A Backpropagation Artificial Neural Network Approach for Loan Status Prediction

Abstract

Providing credit has become the main source of profit for financial and non-financial institutions. However, this transaction might lead to credit risk where debtors are unable to complete their obligations. In this case, the prediction of loan status is extremely important to minimize the risks. The objective of this work is to predict loan status by using a backpropagation algorithm. The used dataset consists of 1 dependent variable and 13 independent variables which 75 % are for data training and 25 % for data testing. There are two main simulation experiments namely simulation involving all predictor variables and another one involving just only predictor variables has a significant relationship with the target variable. The first main simulation experiment shows the best performance metrics from the first model are 94.37% accuracy, 78.57% sensitivity, 98.25% specificity, 91.67% precision, and 84.62% F1 score. The performance metrics of the second one are the same as the best performance metrics of the first simulation. The results of this study can potentially be applied by financial institutions to assist in the feasibility assessment of prospective debtors to reduce company losses.