Fraud detection by profiling aggregate customer anonymous behavior

Patent metadata

Field	Value
Publication number	US-10692058-B2
Application number	US-201715697375-A
Country	US
Kind code	B2
Filing date	Sep 6, 2017
Priority date	Sep 6, 2017
Publication date	Jun 23, 2020
Grant date	Jun 23, 2020

Abstract

Computer implementation methods of processing transactions to determine the fraud risk of transactions incorporating card issuer bin and cardholder location associated with a multitude of customers. The artificial intelligence models developed with such information provide an output of likelihood of fraud for payment card transactions. Disclosed are the methods of utilizing aggregated payment card transaction data at the card issuer bin and card holder location level to improve fraud detection. The implementation of the method is demonstrated to have boosted the performance of the developed models in detection of fraudulent payment cards.

First claim

What is claimed is:

1. A computer implemented method for a computerized fraud detection model, the method comprising:
receiving, by one or more processors via a point of sale computing system, transaction data for a transaction using a transaction card, the transaction data including a primary account number (PAN) of the transaction card, a bank identification number (BIN) of a bank that issued the transaction card, and aggregate geographic information (AGI) associated with the transaction;
generating, by the one or more processors, a PAN profile according to a historical set of transaction data based on the PAN extracted from each transaction data received;
generating, by the one or more processors, a BIN+AGI profile according to the historical set of transaction data based on the BIN and AGI extracted from each transaction data received;
applying, by the one or more processors, the PAN profile and the BIN+AGI profile to a fraud detection model, the fraud detection model being developed from the historical set of transaction data and executed by an artificial intelligence computing system that receives training data from human supervision of the fraud detection model; and
generating, by the one or more processors, a score for the transaction according to the model executed by the artificial intelligence computing system, the score representing a likelihood of fraud of the transaction, based on which the computerized fraud detection model categorizes fraudulent and non-fraudulent activity,
performance of the computerized fraud detection model being enhanced by tracking fraudulent activity at multiple levels including at least one of the transaction card level and an aggregate geographic level.

2. The computer implemented method in accordance with claim 1 , wherein the AGI includes one or more digits of a ZIP code associated with the transaction.

3. The computer implemented method in accordance with claim 1 , further comprising aggregating, by the one or more processors, features of the PAN profile with features related to the BIN+AGI profile.

4. The computer implemented method in accordance with claim 3 , further comprising generating, by the one or more processors, a fraud score based on features associated with the PAN, the BIN, the AGI, or the BIN+AGI profile.

5. The computer implemented method in accordance with claim 1 , further comprising generating, by the one or more processors, favorite merchant(s) variables according to the historical set of transaction data based on a merchant categorical code (MCC), a merchantID, and a location extracted from each transaction data received.

6. The computer implemented method in accordance with claim 1 , further comprising:
tagging, by the one or more processors, each transaction as fraud or non-fraud; and
storing, by the one or more processors, tags of the tagged transactions in the historical set of transaction data.

7. The computer implemented method in accordance with claim 1 , further comprising applying, by the one or more processors, a decay factor to the PAN profile and/or the BIN+AGI profile, the decay factor being event-dependent, time-dependent, and/or distance dependent.

8. A computer program product comprising a non-transitory machine-readable medium storing instructions that, when executed by at least one programmable processor, cause the at least one programmable processor to perform operations comprising
receive, via a point of sale computing system, transaction data for a transaction using a transaction card, the transaction data including a primary account number (PAN) of the transaction card, a bank identification number (BIN) of a bank that issued the transaction card, and aggregate geographic information (AGI) associated with the transaction;
generate a PAN profile according to a historical set of transaction data based on the PAN extracted from each transaction data received;
generate a BIN+AGI profile according to the historical set of transaction data based on the BIN and AGI extracted from each transaction data received;
apply the PAN profile and the BIN+AGI profile to a fraud detection model, the fraud detection model being developed from the historical set of transaction data and executed by an artificial intelligence computing system that receives training data from human supervision of the fraud detection model; and
generate a score for the transaction according to the model executed by the artificial intelligence computing system, the score representing a likelihood of fraud of the transaction based on which the computerized fraud detection model categorizes fraudulent and non-fraudulent activity,
performance of the computerized fraud detection model being enhanced by tracking fraudulent activity at multiple levels including at least one of the transaction card level and an aggregate geographic level.

9. The computer program product in accordance with claim 8 , wherein the AGI includes one or more digits of a ZIP code associated with the transaction.

10. The computer program product in accordance with claim 8 , wherein the operations further comprise aggregate features of the PAN profile with features related to the BIN+AGI profile.

11. The computer program product in accordance with claim 10 , wherein the operations further comprise generate a fraud score based on features associated with the PAN, the BIN, the AGI, or the BIN+AGI profile.

12. The computer program product in accordance with claim 8 , wherein the operations further comprise generate favorite merchant(s) variables according to the historical set of transaction data based on a merchant categorical code (MCC), a merchantID, and a location extracted from each transaction data received.

13. The computer program product in accordance with claim 8 , wherein the operations further comprise:
tag each transaction as fraud or non-fraud; and
store tags of the tagged transactions in the historical set of transaction data.

14. The computer program product in accordance with claim 8 , wherein the operations further comprise apply a decay factor to the PAN profile and/or the BIN+AGI profile, the decay factor being event-dependent, time-dependent, and/or distance dependent.

15. A system comprising:
a programmable processor; and
a machine-readable medium storing instructions that, when executed by the processor, cause the at least one programmable processor to perform operations comprising:
receive, via a point of sale computing system, transaction data for a transaction using a transaction card, the transaction data including a primary account number (PAN) of the transaction card, a bank identification number (BIN) of a bank that issued the transaction card, and aggregate geographic information (AGI) associated with the transaction;
generate a PAN profile according to a historical set of transaction data based on the PAN extracted from each transaction data received;
generate a BIN+AGI profile according to the historical set of transaction data based on the BIN and AGI extracted from each transaction data received;
apply the PAN profile and the BIN+AGI profile to a fraud detection model, the fraud detection model being developed from the historical set of transaction data and executed by an artificial intelligence computing system that receives training data from human supervision of the fraud detection model; and

generate a score for the transaction according to the model executed by the artificial intelligence computing system, the score representing a likelihood of fraud of the transaction based on which the computerized fraud detection model categorizes fraudulent and non-

Assignees

• Fair Isaac Corp

Inventors

• Zoldi Scott Michael
• Xu Heming

Classifications

• G06Q20/383

  Anonymous user system · CPC title

• G06Q20/10 Primary

  specially adapted for electronic funds transfer [EFT] systems; specially adapted for home banking systems · CPC title

• G06Q30/0601

  Electronic shopping [e-shopping] · CPC title

• G06Q20/4016

  involving fraud or risk level assessment in transaction processing · CPC title