Abstract: This disclosure describes techniques for determining whether to approve or deny a borrower's lending-product request by selectively using a heuristic and statistical model. More specifically, a borrower may submit a lending-product request to a Statistical Risk Management (SRM) system, and in doing so the SRM system may analyze relationship attributes of the borrower to determine a likelihood of borrower repaying a loan over a predetermined time period, and avoid being charged off. In some examples, the SRM system may execute a plurality of statistical models to determine a charge-off probability score. Each statistical model may be based on a set, or subset of historical lending-product data. A subset of historical lending-product data may be based on a selection bias of shared characteristics within a set of historical lending-product data. The selection bias may be based on characteristics of a lending-product request or relationship attributes of a borrower.

Abstract: This disclosure describes techniques for determining whether to approve or deny a borrower's lending-product request by selectively using a heuristic and statistical model. More specifically, a borrower may submit a lending-product request to a Heuristic-Statistical Risk Management (HS-RM) system, and in doing so the HS-RM system may analyze relationship attributes of the borrower to determine a likelihood of borrower repaying a loan over a predetermined time period, and avoid being charged off. In some examples, the HS-RM system may execute a plurality of statistical models to determine a charge-off probability score. Each statistical model may be based on a set, or subset of historical lending-product data. A subset of historical lending-product data may be based on a selection bias of shared characteristics within a set of historical lending-product data. The selection bias may be based on characteristics of a lending-product request or relationship attributes of a borrower.