US 11,810,009 B2
Methods of explaining an individual predictions made by predictive processes and/or predictive models
Gregory Dean Jorstad, Draper, UT (US); and Thomas Patrick Prendergast, Jr, Draper, UT (US)
Assigned to Synchrony Bank, Stamford, CT (US)
Filed by Synchrony Bank, Stamford, CT (US)
Filed on Sep. 7, 2022, as Appl. No. 17/930,076.
Application 17/930,076 is a continuation of application No. 16/293,407, filed on Mar. 5, 2019, granted, now 11,475,322.
Prior Publication US 2023/0109251 A1, Apr. 6, 2023
Int. Cl. G06N 5/04 (2023.01); G06N 20/00 (2019.01); G06F 3/14 (2006.01)
CPC G06N 5/04 (2013.01) [G06N 20/00 (2019.01); G06F 3/14 (2013.01)] 21 Claims
OG exemplary drawing
 
1. A computer-implemented method comprising:
receiving an input record on an explanation computing device, wherein the explanation computing device includes an input message broker, a model executing engine, one or more sample bins, an explanation live process, and an output message broker, wherein the input record includes a set of input variables and a set of values corresponding to the set of input variables, and wherein the input record is received through the input message broker;
generating an actual prediction by processing the input record using the model executing engine;
generating a set of modified input records, wherein the set of modified input records are generated by the explanation computing device, wherein a modified input record is generated by modifying a particular value associated with a particular input variable without modifying other values associated with other input variables, and wherein the particular value is modified using a set of sample values obtained from the one or more sample bins;
generating a set of sample predictions by processing the set of modified input records using the model executing engine;
executing the explanation live process to evaluate the set of sample predictions against the actual prediction to identify an influential input variable; and
providing a justification for identification of the influential input variable through the output message broker, wherein the justification includes sample predictions corresponding to changes to influential input variable values.