SAS' advanced analytical techniques have a proven ability to quickly and accurately forecast the risk of credit losses at financial institutions. It is able to provide the answers to questions such as "Which applicants should be accepted or rejected?", "Which accounts are likely to go into arrears?", and 'Which of the customers in arrears will pay?". This paper is intended as a primer to the application of data mining techniques available in SAS/Enterprise Miner™ to the credit scoring process so as to minimise the risk of delinquency-Credit scoring is a method of quantifying the risk of a particular credit applicant. The final score of an applicant is obtained from the sum of the individual scores that are based on a number of different characteristics such as demographics, employment information and debt-to-income ratios. The score classifies the applicant into a particular good/bad odds group. This grouping is then compared to a pre-defined cut-off point to determine the risk level of the applicant.
The underlying assumption of the aforementioned process is that past behaviour accurately reflects future behaviour. Inductive models such as logistic regression, neural networks and decision trees can be used to infer patterns and relationships from historical credit data and generalise these findings to score new applicants. A high-level explanation of these techniques is provided and their characteristics compared. A brief overview of the reject inference problem is also covered.