Methods and apparatus for payment fraud detection

Patent metadata

Field	Value
Publication number	US-11494775-B2
Application number	US-201916560764-A
Country	US
Kind code	B2
Filing date	Sep 4, 2019
Priority date	Sep 4, 2019
Publication date	Nov 8, 2022
Grant date	Nov 8, 2022

Abstract

This application relates to apparatus and methods for identifying fraudulent transactions. In some examples, a computing device generates a decision matrix to identify fraudulent transactions. To generate the decision matrix, the computing device may determine scores for a plurality of transactions, and may determine transaction categories for each transaction based on the scores. The computing device may also determine a number of predictable features based on applying machine learning techniques to the transactions. A risk category is then determined for the number of predictable features. The computing device generates the decision matrix based on the transaction categories and the risk categories. In some examples, the computing device applies the generated decision matrix to an ongoing purchase transaction to determine if the ongoing purchase transaction is fraudulent. In some examples, the computing device prevents completion of the purchase transaction if the purchase transaction is determined to be fraudulent.

First claim

What is claimed is:

1. A system comprising:
a computing device configured to:
receive purchase data from a first device over a communication network, wherein the received purchase data comprises information associated with at least one purchase transaction;
generate at least a first value based on the received purchased data, wherein the at least first value comprises information about whether the at least one purchase transaction is fraudulent;
determine a value category for the at least one purchase transaction based on the at least first value, wherein determining the value category further comprises:
applying, using a first machine learning process, each of a plurality of models to the at least one purchase transaction to generate a model score for each of the plurality of models;
determining a model category for each of the plurality of models based on the corresponding model score; and
determining the value category for the at least one purchase transaction based on the determined model categories;

determine a plurality of features associated with the first value based on applying a second machine learning process to the purchase data and the first value, wherein each feature of the plurality of features is associated with at least a second value;
determine a risk category for each of the plurality of features based on applying at least one binning process to the at least second value;
generate decision data based on the determined value category and the determined risk categories associated with the plurality of features, wherein the decision data comprises information associated with a plurality of conditions for determining whether the at least one purchase transaction is fraudulent;
determine whether the at least one purchase transaction is fraudulent based on the decision data;
generate transaction allowability data associated with an allowance of the at least one purchase transaction in response to determining that the at least one purchase transaction is not fraudulent based on the decision data; or
generate transaction allowability data associated with a rejection of the at least one purchase transaction in response to determining that the at least one purchase transaction is fraudulent based on the decision data; and
transmit the transaction allowability data to the first device in response to the received purchase data.

2. The system of claim 1 , wherein determining whether the at least one purchase transaction is fraudulent comprises comparing the decision data to the determined value category and the determined risk categories.

3. The system of claim 1 , wherein the plurality of conditions comprises:
at least one condition that the at least one purchase transaction is associated with at least one value category; and
at least one condition that the at least one purchase transaction is associated with at least one risk category.

4. The system of claim 1 , wherein determining whether the at least one purchase transaction is fraudulent comprises determining whether the at least one purchase transaction is to be allowed or denied.

5. The system of claim 1 , wherein determining whether the transaction is fraudulent comprises determining whether the at least one purchase transaction is to be reviewed.

6. The system of claim 5 , wherein the computing device is further configured to:
transmit a review request message to a second computing device, wherein the review request message identifies the at least one purchase transaction; and
receive, in response to the review request message, a review response message identifying whether the at least one purchase transaction is to be allowed or denied.

7. The system of claim 1 , wherein determining the plurality of features comprises applying at least one machine learning process to the at least one transaction and the at least second value.

8. The system of claim 7 , wherein the at least one machine learning process was trained with historical purchase transactions.

9. The system of claim 1 , wherein the computing device is configured to determine customer history data associated with the at least one purchase transaction, wherein determining the value category for the at least one purchase transaction is based on the associated customer history data.

10. A method comprising:
receiving, by a computing device, purchase data from a first device over a communication network, wherein the received purchase data comprises information associated with at least one purchase transaction;
generate, by the computing device, at least a first value based on the received purchased data, wherein the at least first value comprises information about whether the at least one purchase transaction is fraudulent;
determining, by the computing device, a value category for the at least one purchase transaction based on the at least first value wherein determining the value category further comprises:
applying, using a first machine learning process, each of a plurality of models to the at least one purchase transaction to generate a model score for each of the plurality of models;
determining a model category for each of the plurality of models based on the corresponding model score; and
determining the value category for the at least one purchase transaction based on the determined model categories;

determining, by the computing device, a plurality of features associated with the first value based on applying a second machine learning process to the purchase data and the first value, wherein each feature of the plurality of features is associated with at least a second value;
determining, by the computing device, a risk category for each of the plurality of features based on applying at least one binning process to the at least second value;
generating, by the computing device, decision data based on the determined value category and the determined risk categories, wherein the decision data comprises information associated with a plurality of conditions for determining whether the at least one purchase transaction is fraudulent;
determining, by the computing device, whether the at least one purchase transaction is fraudulent based on the decision data;
generating, by the computing device, transaction allowability data associated with an allowance of the at least one purchase transaction in response to determining that the at least one purchase transaction is not fraudulent based on the decision data;
and
transmitting, by the computing device, the transaction allowability data to the first device in response to the received purchase data.

11. The method of claim 10 , wherein determining whether the at least one purchase transaction is fraudulent comprises comparing the decision data to the determined value category and the determined risk categories.

12. The method of claim 10 , wherein the plurality of conditions comprises:
at least one condition that the at least one purchase transaction is associated with at least one value category; and
at least one condition that the at least one purchase transaction is associated with at least one risk category.

13. The method of claim 10 , wherein determining the plurality of features comprises applying at least one machine learning process to the at least one transaction and the at least second value.

14. A non-transitory computer readable medium having instructions stored thereon, wherein the instructions, when executed by at least one processor of a computing device, cause the computing device to perform operations comprising:
receiving purchase data from a first device over a communication network, wherein the received pur

Assignees

• Walmart Apollo Llc

Inventors

• Kang Linhong
• Matsumoto Kazuo
• Thangamani Premalatha
• Xia Jing
• Gagpalliwar Saurabh Vivek

Classifications

• G06Q20/4016 Primary

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

• G06Q20/389

  Keeping log of transactions for guaranteeing non-repudiation of a transaction · CPC title

• G06Q20/405

  Establishing or using transaction specific rules · CPC title

• G06N20/00

  Machine learning · CPC title

• G06Q30/0609 Primary

  Qualifying participants for shopping transactions (payment transaction verification G06Q20/401) · CPC title