Automatic adjustment of limits based on machine learning forecasting

Inventors

Chen, Bryant • Xu, Lillian • Jin, Jeanette

Assignees

Brex Inc

Abstract

There are provided systems and methods for automatic adjustment of limits based on machine learning forecasting. An entity, such as company or other entity, may purchase items utilizing a payment instrument or card provided to the company by a credit provider system or entity. In order to provide proper underwriting for credit extensions, such as balances and limits of extendable credit, the credit provider system may utilize a forecasting machine learning (ML) model trained to predict a future global balance of funds or a likelihood of repayment of the extended credit limit. This may be based on information retrievable balances from a banking system and a staleness of this data. When the data is stale and has not been updated, the forecasted balance may have a wider range, and thus risk factors may designate less risky and lower limits.

Core Innovation

The invention provides a credit provider system and an underwriting system for automatically adjusting available credit for electronic transaction processing for an entity. A machine learning (ML) model predicts predictive scores for delinquency and/or provides a balance forecast for an entity at a closing time for a billing cycle associated with the available credit. The model uses features extractable from a global balance of funds and an account staleness factor associated with a time since a last update of the global balance.

The system dynamically adjusts the available credit based on the balance forecast and/or the predictive score, where the predictive score is based on one or more risk factors for the ML model. A predictive confidence interval is used for deciding credit adjustments, including conditions where the recalculated predictive score falls outside a predictive confidence interval. The approach includes detecting usage of the available credit over a period of time, and retraining the ML model based on the detected usage, including modifying values for nodes and adjusting the ML model for performance improvement.

The invention further includes receiving an update to the global balance of funds, updating the account staleness factor to a current time of receiving the update, recalculating the predictive score or balance forecast using the updated features, and dynamically readjusting the available credit based on the recalculated predictive score. The ML model includes plurality of ML model layers with one or more nodes trained using training data from a plurality of entities, including tree-based algorithm training, and the disclosed processing supports end-to-end model training, deployment, prediction at billing cycle closing time, and iterative recalculation and adjustment as new global balance information is received.

Claims Coverage

The partial content provides three independent claim sets (system, method, and non-transitory machine-readable medium). Each includes an ML-based prediction using global balance and an account staleness factor, followed by dynamic credit adjustment, with retraining based on detected credit usage and recalculation/realignment when an updated global balance is received.

Across the independent claims, the inventive core is an ML model trained with tree-based algorithm training to predict delinquency and/or generate a balance forecast using features from a global balance and an account staleness factor, followed by dynamic adjustment of available credit. The claims further require detecting available-credit usage, retraining based on detected usage/feedback, and recalculating and readjusting available credit when updated global balance information is received, with predictive confidence interval functionality appearing in the method and machine-readable medium independent claim sets.

Stated Advantages

Documented Applications