Title: Predicting bank performance using machine learning: a case of troubled banks in India

Authors: Sumedha Tuteja; Punam Bhoyar; Krishna Kumar Singh; Aruna Dev Rroy

Addresses: Indira Institute of Management, Pune, 85/5 A, Tapasya, New Mumbai Pune Highway, Tathawade, Pune, Maharashtra 411033, India ' Indira Institute of Management, Pune, 85/5 A, Tapasya, New Mumbai Pune Highway, Tathawade, Pune, Maharashtra 411033, India ' Symbiosis International University (SIU), SCIT, Plot No. 15, Rajiv Gandhi Infotech Park, MIDC, Hinjewadi, Phase 1, Pune, Maharashtra 411057, India ' Royal Global University, Betkuchi, Opp. Tirupati Balaji Temple, ISBT, NH-37, Guwahati-35, Dist-Kamrup (Metro), Assam, India

Abstract: The significance of a bank's financial stability for achieving a thriving and robust economy cannot be overstated. This study develops a predictive model using machine learning (ML) to categorise banks into low- or high-performance. The non-performing assets (NPA) levels of the Indian banking sector, notably for the public sector banks (PSBs), have increased significantly since 2015. Hence, the model has been created using bank performance data of PSBs, specifically for the period 2015-2020. The authors first identified logical groups by using the unsupervised K-means clustering method; and, subsequently, deployed a supervised algorithm for prediction: the classification and regression tree (CART). The model has an overall prediction accuracy of 0.9375, a sensitivity of 0.8571 and a specificity of 0.9600. This study is unique since it uses only data from banks with high NPA levels to create a predictive model for bank performance.

Keywords: banks; bank performance; clustering method; decision tree method; financial performance; India; machine learning; non-performing loans; public sector banks; predictive modelling.

DOI: 10.1504/IJPEE.2023.133638

International Journal of Pluralism and Economics Education, 2023 Vol.14 No.1, pp.55 - 72

Received: 16 Feb 2023
Accepted: 20 May 2023
Published online: 26 Sep 2023 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article