# US 12,182,819 B1 · Fraud detection using augmented analytics > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,182,819 B1 - Document ID: US12182819B1 - Status: Granted - Kind: B1 - Application number: 18/149,109 - Filed: 2023-01-01 - Granted: 2024-12-31 - Earliest claimed: 2019-09-30 - Assignees: United Services Automobile Association (USAA) - Applicants: United Services Automobile Association (USAA) - Inventors: Gregory D. Hansen; Andre R. Buentello; Ashley R. Philbrick; Jose L. Romero, Jr.; Reynaldo Medina, III; Curtis M. Bell; Yevgeniy Khmelev; Ruthie D. Lyle; Victor Kwak; Jon D. McEachron; Stacy C. Huggar - Web page: https://patentgrep.com/patent/US12182819B1 ## CPC - G06Q 20/4016: {involving fraud or risk level assessment in transaction processing} - G06N 20/00: Machine learning - G06Q 20/405: {Establishing or using transaction specific rules} - G06Q: INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS ## Abstract Systems and methods ingest extensive data regarding real-time transactions to determine correlation between a variety of parameters and possible fraud events. Possible fraud can be predicted, and notifications delivered to a fraud team. Correlations can be discovered using unsupervised machine learning, and particular systems and methods can utilize natural language processing to both receive and provide information concerning fraud risk and countermeasures. ## Claims ### Claim 1 A method, comprising: receiving fraud contextual data describing context surrounding fraud events and transaction data describing transactional contextual data associated with transactions; determining a plurality of correlations between the fraud contextual data and the transactional contextual data using unsupervised machine learning; analyzing the plurality of correlations using supervised machine learning to identify at least one causative correlation; receiving real-time transaction data, wherein the real-time transaction data includes data describing a requested transaction and real-time transactional contextual data associated with the requested transaction; determining whether a fraud correlation match exists for the requested transaction based on the real-time transaction data corresponding to at least one of the at least one causative correlation; and causing an action to close the real-time transaction associated with the real-time transaction data based on the fraud correlation match, wherein closing the real-time transaction includes causing completion of the real-time transaction when the correlation match does not exist, and wherein closing the real-time transaction includes freezing the real-time transaction when the fraud correlation match exists. ### Claim 2 (depends on 1) The method of claim 1, comprising: providing a notification that the real-time transaction has been frozen. ### Claim 3 (depends on 1) The method of claim 1, comprising: determining a context difference between the fraud contextual data and the real-time transaction; and determining an action to modify the context difference. ### Claim 4 (depends on 1) The method of claim 1, comprising: preparing at least one of the fraud data and the real-time transaction data according to a record format, wherein the fraud contextual data respectively associated with each of the fraud events is ingested to an individual record according to the record format. ### Claim 5 (depends on 4) The method of claim 4, wherein machine learning identifies one or more fields of the record format to be populated by the fraud data and the real-time transaction data, and wherein the record format is dynamically updated based on identification of a new field type by the machine learning. ### Claim 6 (depends on 1) The method of claim 1, comprising: discarding, using supervised machine learning, at least one of the plurality of correlations based on the at least one of the plurality of correlations being non-causative. ### Claim 7 (depends on 1) The method of claim 1, comprising: causing an incentive to be offered to a party associated with the requested transaction based on the party completing a behavior that eliminates the fraud correlation match. ### Claim 8 A system, comprising a non-transitory computer-readable medium storing instructions configured to effectuate: a data ingestion component configured to receive fraud contextual data describing context surrounding fraud events, transaction data describing transactional contextual data associated with transactions, and real-time transaction data, wherein the real-time transaction data includes data describing a requested transaction and real-time transactional contextual data associated with the requested transaction; an unsupervised machine learning component configured to determine a plurality of correlations between the fraud contextual data and the transactional contextual data using unsupervised machine learning; a supervised machine learning component configured to analyze the plurality of correlations using supervised machine learning to identify at least one causative correlation; a correlation component configured to determine whether a fraud correlation match exists for the requested transaction based on the real-time transaction data corresponding to at least one of the at least one causative correlation; and a user interface configured to provide a notification based on the correlation. ### Claim 9 (depends on 8) The system of claim 8, wherein the correlation component is configured to cause an action to close the real-time transaction associated with the real-time transaction data based on the fraud correlation match, wherein closing the real-time transaction includes causing completion of the real-time transaction when the fraud correlation match does not exist, and wherein closing the real-time transaction includes freezing the real-time transaction when the fraud correlation match exists. ### Claim 10 (depends on 9) The system of claim 9, wherein the instructions are further configured to effectuate: a transaction approval component configured to complete the real-time transaction or freeze the real-time transaction based on the fraud correlation match. ### Claim 11 (depends on 8) The system of claim 8, wherein the instructions are further configured to effectuate: a data preparation component configured to prepare at least one of the fraud data and the real-time transaction data according to a record format, wherein the fraud contextual data respectively associated with each of the fraud events is ingested to an individual record according to the record format. ### Claim 12 (depends on 8) The system of claim 8, wherein the instructions are further configured to effectuate: a fraud detection database. ### Claim 13 (depends on 8) The system of claim 8, wherein the correlation component is configured to determine a context difference between the fraud contextual data and the real-time transaction. ### Claim 14 (depends on 13) The system of claim 13, wherein the correlation component is configured to determine an action to modify the context difference. ### Claim 15 A non-transitory computer-readable medium storing instructions that when executed by a processor effectuate operations comprising: receiving fraud contextual data describing context surrounding fraud events and transaction data describing transactional contextual data associated with transactions; determining a plurality of correlations between the fraud contextual data and the transactional contextual data using unsupervised machine learning; analyzing the plurality of correlations using supervised machine learning to identify at least one causative correlation; receiving real-time transaction data, wherein the real-time transaction data includes data describing a requested transaction and real-time transactional contextual data associated with the requested transaction; determining whether a fraud correlation match exists for the requested transaction based on the real-time transaction data corresponding to at least one of the at least one causative correlation; and causing an action to close the real-time transaction associated with the real-time transaction data based on the fraud correlation match, wherein closing the real-time transaction includes causing completion of the real-time transaction when the fraud correlation match does not exist, and wherein closing the real-time transaction includes freezing the real-time transaction when the fraud correlation match exists. ### Claim 16 (depends on 15) The non-transitory computer-readable medium of claim 15, wherein the instructions when executed effectuate operations comprising: providing a notification that the real-time transaction has been frozen. ### Claim 17 (depends on 15) The non-transitory computer-readable medium of claim 15, wherein the instructions when executed effectuate operations comprising: determining a context difference between the fraud contextual data and the real-time transaction. ### Claim 18 (depends on 17) The non-transitory computer-readable medium of claim 17, wherein the instructions when executed effectuate operations comprising: determining an action to modify the context difference. ### Claim 19 (depends on 15) The non-transitory computer-readable medium of claim 15, wherein machine learning identifies one or more fields of a record format to be populated by the fraud data and the real-time transaction data, wherein the fraud contextual data respectively associated with each of the fraud events is ingested to an individual record according to the record format. ### Claim 20 (depends on 19) The non-transitory computer-readable medium of claim 19, wherein the record format is dynamically updated based on identification of a new field type by the machine learning.