Abstract:
In the Sri Lankan context, microfinance is the main banking service provider for the 'unbankable' people
who live in rural areas. Due to the informal nature and high-interest rate, microfinance continues to
operate in a falling field. In addition, in the appraisal stage, rural borrowers tend to make mistakes
while filling out loan applications causing default situations. As a result, microfinance institutions
demotivate and people shifting away from traditional forms of borrowing to informal forms is socially
problematic. Therefore, in order to address those issues, the researchers collected primary data
through ABC bank and analyzed data relating to 10000 random borrowers' records and set of loan
forms. The purpose of this research is to automate paper-based loan applications to recognize
handwritten characters with the use of information technology through image processing techniques
and analyze customers' determinant factors to predict interest rates for microfinance facilities. The
finding of our study shows that Regression analysis models can gain the best result in predicting the
interest rate. Together Ridge regression analysis and XGBoost regression models gave the most
accurate results compared to other models in interest rate prediction. By using this interest rate
prediction, microfinance institutions can offer a suitable interest rate that is convenient for the loan
borrower. The application automation reduces the paperwork and the manual effort needed to process
and increases data accuracy. The developed model can enhance the repayment performance of
microfinance firms and prevent defaults by borrowers at the loan appraisal stage.
