Implementation of Early Warning System of Loan Customer Default for Bank in Sri Lanka

Abstract

Credit has been the main source of income for bank throughout banking history. With deteriorating economic conditions, identifying the best person to grant credit has become a challenge. In the current Sri Lankan context, most banks are facing the issue of provisioning for bad loans which remove a proportion of profit. Thus, it directly affects the profitability of the bank and increases credit risk. This project aims to identify a high-risk customer after granting a loan. A customer could be a good customer in the past and now looking for a new credit facility. Further, the revaluation of the customer base is done periodically by banks to identify the position of customers in the ever-changing environment. The model built in this research will ease the customer revaluation process and provide a window of early warning of customer default. For this purpose, data science will be used that will aid algorithms like K means, ad boost, Logistic regression, random forest, and neural networks. The analysis will be done using easy to implement a tool called Orange. Predicting customers who are going to default will have a lot of plus points to the bank and its customers. Hence now the bank can play an advisory role to its customers and help them to overcome a future disaster. In a changing environment, banks cannot just be a lender; banks are now moving into an advisory role as well.