	US 12,236,439 B2
	Reducing false positives using customer feedback and machine learning
Timothy Kramme, Parker, TX (US); Elizabeth A. Flowers, Bloomington, IL (US); Reena Batra, Alpharetta, GA (US); Miriam Valero, Bloomington, IL (US); Puneit Dua, Bloomington, IL (US); Shanna L. Phillips, Bloomington, IL (US); Russell Ruestman, Minonk, IL (US); and Bradley A. Craig, Normal, IL (US)
Assigned to State Farm Mutual Automobile Insurance Company, Bloomington, IL (US)
Filed by State Farm Mutual Automobile Insurance Company, Bloomington, IL (US)
Filed on Jun. 7, 2023, as Appl. No. 18/207,069.
Application 18/207,069 is a continuation of application No. 17/080,476, filed on Oct. 26, 2020, granted, now 11,687,938.
Application 17/080,476 is a continuation of application No. 15/465,832, filed on Mar. 22, 2017, granted, now 10,872,339, issued on Dec. 22, 2020.
Claims priority of provisional application 62/365,699, filed on Jul. 22, 2016.
Claims priority of provisional application 62/331,530, filed on May 4, 2016.
Claims priority of provisional application 62/318,423, filed on Apr. 5, 2016.
Claims priority of provisional application 62/313,196, filed on Mar. 25, 2016.
Prior Publication US 2023/0316285 A1, Oct. 5, 2023
Int. Cl. G06Q 30/02 (2023.01); G06N 5/046 (2023.01); G06N 20/00 (2019.01); G06Q 20/10 (2012.01); G06Q 20/20 (2012.01); G06Q 20/24 (2012.01); G06Q 20/32 (2012.01); G06Q 20/34 (2012.01); G06Q 20/40 (2012.01); G06Q 30/018 (2023.01); G06Q 30/0207 (2023.01); G06V 30/194 (2022.01); G06V 30/41 (2022.01); G06Q 30/0241 (2023.01)

CPC G06Q 30/0185 (2013.01) [G06N 5/046 (2013.01); G06N 20/00 (2019.01); G06Q 20/102 (2013.01); G06Q 20/20 (2013.01); G06Q 20/24 (2013.01); G06Q 20/3224 (2013.01); G06Q 20/34 (2013.01); G06Q 20/401 (2013.01); G06Q 20/4016 (2013.01); G06Q 20/407 (2013.01); G06Q 20/409 (2013.01); G06Q 30/0225 (2013.01); G06V 30/194 (2022.01); G06V 30/41 (2022.01); G06Q 30/0248 (2013.01)]

18 Claims

1. A computer-implemented method of using a machine-learned model to determine a cause of a false positive fraud alert, the method comprising:

receiving, by one or more processors, transaction data associated with a financial transaction initiated by a customer;

determining, using a rules engine applying one or more fraud detection rules, a fraud alert associated with the financial transaction;

receiving, by the one or more processors and from a customer computing device, customer feedback indicating that the fraud alert is a false positive fraud alert;

in response to receiving the customer feedback, providing the transaction data as input to a machine-learned model trained to output a piece of data within the transaction data that caused the false positive fraud alert;

determining, based on the output of the machine-learned model, a cause associated with the false positive fraud alert; and

modifying, based on the cause, a first fraud detection rule of the rules engine including a threshold associated with the piece of data that caused the false positive fraud alert.