COMPARISON OF THE PERFORMANCE OF SEVERAL DATA MINING TECHNIQUES FOR LOAN-GRANTING DECISIONS

摘要

Assessing financial status of customers and their credit worthiness, evaluating loan-granting policies, and loan payment prediction are critical to the business of banks and money lending institutions. The paper compares the classification accuracy of three data mining techniques such as decision trees, neural networks, and logit regression for loan payment prediction. The initial simulation results show that the difference in the classification accuracy between the three methods appears to be insignificant. We recommend using the three techniques in tandem and using the majority voting to achieve a more reliable decision. The decision tree approach might be preferable to the other ones because it can produce understandable rules that allow one to explain the rationale behind the decision.