Classification of non-performing financing using logistic regression and synthetic minority over-sampling technique-nominal continuous (SMOTE-NC)

Abstract

Financing analysis is the process of analyzing the ability of bank customers to pay installments to minimize the risk of a customer not paying installments, which is also called Non-Performing Financing (NPF). In 2020 the NPF ratio at one of the Islamic banks in Indonesia increased due to the decline in people’s income during the Covid-19 pandemic. This phenomenon has led to bad banking performance. In December 2020 the percentage of NPF was 17%. The imbalance between the number of good-financing and NPF customers has resulted in poor classification accuracy results. Therefore, this study classifies NPF customers using the Logistic Regression and Synthetic Minority Over-sampling Technique Nominal Continuous (SMOTE-NC) method. The results of this study indicate that the logistic regression with SMOTE-NC model is the best model for the classification of NPF customers compared to the logistic regression method without SMOTE-NC. The variables that have a significant effect are financing period, type of use, type of collateral, and occupation. The logistic regression with SMOTE-NC can handle the imbalanced dataset and increase the specificity when using logistic regression without SMOTE-NC from 0.04 to 0.21, with an accuracy of 0.81, sensitivity of 0.94, and precision of 0.86.