This study aims to investigate the potential of various machine-learning models to expedite the credit rating assignment process for non-banking finance companies (NBFCs) in India. It seeks to address the limitations of traditional credit rating agencies, which often update ratings only after an organization has begun defaulting, by providing timely information for proactive decision-making.
This study evaluates six machine learning models – support vector machine, artificial neural network, naïve Bayes, decision tree, random forest and gradient boosting decision tree – using a data set of top 50 Indian NBFCs. The data span six fiscal years, from 2014–2015 to 2019–2020, specifically focusing on the pre-crisis period of COVID-19 and were primarily sourced from the Center for Monitoring Indian Economy database, with missing data supplemented by annual reports of the listed companies.
The results indicate that the gradient boosting decision tree algorithm outperforms the other models in predicting credit ratings, followed by the polynomial kernel support vector machine and random forest algorithms. This suggests that machine-learning models, particularly gradient boosting, can provide more efficient and accurate credit rating predictions for NBFCs.
The findings of this study have practical implications for NBFCs and rating agencies. By incorporating machine learning models, the credit rating process can be significantly expedited, offering timely insights for financial institutions and regulators to implement proactive measures to mitigate risk.
This study contributes to the existing literature by applying and comparing machine learning techniques to predict credit ratings, specifically for Indian NBFCs during pre-crisis and crisis period of COVID-19, a sector that plays a vital role in the country’s financial ecosystem. This study elucidates the potential of modern machine learning models to enhance the timeliness and accuracy of credit rating assessments in this context.
