THE IMPACT OF AI-BASED CREDIT SCORING MODELS ON REDUCING LOAN DEFAULT RATES IN COMMERCIAL BANKS

Abstract

This study examines how AI- Based credit scoring models are efficient in terms of decreasing the default rates of loans than conventional systems used by commercial banks in Pakistan. Historical credit data is the main basis of traditional credit scoring models, e.g. FICO, and is ineffective with borrowers who have low or no credit history. These are also the models that are not dynamic, i.e. change according to the real time changes in the financial behavior of a borrower. Conversely, AI-based models can use a wider dataset including history of transactions, utility payments and even social media activity and give a more comprehensive picture of how financially healthy a borrower is. With the inclusion of diverse data types, AI models can provide a more precise evaluation of creditworthiness, particularly among people who are not directly connected to the main credit system. The research functions as an analysis of the data on five commercial banks in Lahore, Pakistan, to determine the rates of loans default and the accuracy of prediction of the models based on AI and traditional ones. The findings indicate that AI-based models minimize the loan default rates as the average rate of loan default is 7 per cent in comparison to 13 per cent with conventional models. The AI models are also highly predictive because it uses dynamic and real-time data which the traditional models never consider. Moreover, it was found that there is a strong negative relationship between borrower income and loan default rates with the higher-income borrowers having less possibility to default. These results can demonstrate the promise of AI-based credit scoring models in enhancing predictions of loan default, enhancing financial inclusion, and creating more accurate risk assessments of credit risk.