公开(公告)号：US11995714B1

公开(公告)日：2024-05-28

申请号：US17475215

申请日：2021-09-14

Applicant: STRIPE, INC.

Inventor： Mengjie Ding ,  Bo Xu ,  Yaakov Erlichman

IPC: G06Q40/03 ,  G06Q40/06

CPC classification number: G06Q40/03 ,  G06Q40/06

Abstract: Systems and methods for implementing a machine learning loan portfolio loss forecasting system are provided. A current state of active loans of a loan portfolio during a first time period of a set of time periods may be determined. For each of the set of time periods, a roll rate of the active loans from each delinquency state to a subsequent delinquency state may be determined based on historical data of the loan portfolio. The machine learning model may then, iteratively, for each subsequent time period, determine a percentage of the active loans that will transition to each of the set of delinquency states during the subsequent time period based on a current state of the active loans during a previous time period and the roll rate from each delinquency state to a subsequent delinquency state for the subsequent time period.