Rule Induction Methods For Credit Scoring

Abstract

Credit scoring is the term used by the credit industry to describe methods used for classifying applicants for credit into risk classes according to their likely repayment behavior (e.g. “default” and “non-default”).  The credit industry has been using such methods as logistic regression, discriminant analysis, and various machine learning techniques to more precisely identify creditworthy applicants who are granted credit, and non-creditworthy applicants who are denied credit.  Accurate classification is of benefit both to the creditor (in terms of increased profit or reduced loss) and to the loan applicant (avoiding overcommitment).  This paper examines historical data from consumer loans issued by a financial institution to individuals that the financial institution deemed to be qualified customers.  The data set consists of the financial attributes of each customer and includes a mixture of loans that the customers paid off or defaulted upon.  The paper uses rule induction methods (decision trees) to predict whether a particular applicant paid off or defaulted upon his/her loan.  The main advantage of decision trees is their ability to generate if-then classification rules which are intuitive and easy to understand. Rules could be explained to business managers who would need to approve their implementation as well as loan applicants as the reason for denying a loan.  The paper compares the correct classification accuracy rates of several decision tree algorithms with other data mining methods proposed in earlier works.