Abstract:
Loan is one of the furthermost central products of the financial associations. Finance
providers can leverage machine learning and artificial intelligence techniques to enhance
efficiency and improve customer service. In Oromia Micro Financing, stockholders
provide loans to borrowers in exchange for the promise of repayment with interest. The
repayment history of the loaners provides a way to analysis potential customers to the
financing institute by the processes of evaluation and prediction. If proper repayment by
the borrower is good, then the lender would make profit from the interest. However, if the
borrower fails to repay the loan, then the lender loses money. Moreover, loan payment to
the new customers can be decided by some similarity pattern of the previous customers
who have some relevant features of the new customer. In this study, dataset from Oromia
Saving and Credit Association (OSCA)-WALQO was used to train KNN, SVM, MLP
models and two ensemble approaches (stacking and blended) were developed to
determine if the borrower has the ability to repay its loan and deciding the sanction of
loan to the new customer. Features that were determinant in improving classification and
prediction of the models were identified. The blended ensemble approach has shown
higher accuracy in classifying as defaulter (97%) and non-defaulter (98%) than the other
approach.
