SYSTEMS AND METHODS FOR INCORPORATING LONG-TERM PATTERNS IN ONLINE FRAUD DETECTION

• US 20180218283A1
• Filed: 01/27/2017
• Published: 08/02/2018
• Est. Priority Date: 01/27/2017
• Status: Active Grant

• Current Assignee: Meta Platforms, Inc. (f/k/a Facebook, Inc.)
• Original Assignee: Meta Platforms, Inc. (f/k/a Facebook, Inc.)
• Inventors: Jenson, Daniel Adam
• Granted Patent: US 10,990,896 B2
• Time in Patent Office: Days
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06Q 30/0185 Product, service or business identity fraud
• CPC Class Codes: G06Q 50/01 Social networking

Abstract

Systems, methods, and non-transitory computer readable media can generate one or more first machine learning models, where each of the one or more first machine learning models is associated with a respective portion of a first period of time. A second machine learning model incorporating the one or more first machine learning models as features can be generated, where the second machine learning model is associated with a second period of time. A respective weight associated with each of the one or more first machine learning models can be determined. It can be determined whether a content item is associated with a category based on the second machine learning model.

Claims

1. A computer-implemented method comprising:

generating, by a computing system, one or more first machine learning models, each of the one or more first machine learning models associated with a respective portion of a first period of time;

generating, by the computing system, a second machine learning model incorporating the one or more first machine learning models as features, the second machine learning model associated with a second period of time;

determining, by the computing system, a respective weight associated with each of the one or more first machine learning models; and

determining, by the computing system, whether a content item is associated with a category based on the second machine learning model.

2. The computer-implemented method of claim 1, wherein the first period of time is divided into a plurality of portions based on a unit of time.

3. The computer-implemented method of claim 2, wherein the first period of time is a year, the unit of time is a month, and the second period of time is a week.

4. The computer-implemented method of claim 2, wherein each of the one or more first machine learning models is trained based on respective first training data including content items associated with the respective portion of the first period of time.

5. The computer-implemented method of claim 4, wherein the second machine learning model is trained based on second training data including content items associated with the second period of time.

6. The computer-implemented method of claim 5, wherein each of the one or more first machine learning models generates a score for a first content item in the second training data.

7. The computer-implemented method of claim 6, wherein the second training data indicates a score assigned to the first content item, and wherein the respective weight is determined based on a comparison of the score assigned to the first content item and the scores generated by the one or more first machine learning models for the first content item.

8. The computer-implemented method of claim 1, wherein the category relates to online fraud and a pattern associated with the category is cyclical over the first period of time.

9. The computer-implemented method of claim 1, wherein features associated with the one or more first machine learning models include one or more of:

detection-based features or memory-based features.

10. The computer-implemented method of claim 1, wherein features associated with the one or more first machine learning models include one or more of:

numeric features, text features, or image features.

11. A system comprising:

at least one hardware processor; and

a memory storing instructions that, when executed by the at least one processor, cause the system to perform;

generating one or more first machine learning models, each of the one or more first machine learning models associated with a respective portion of a first period of time;

generating a second machine learning model incorporating the one or more first machine learning models as features, the second machine learning model associated with a second period of time;

determining a respective weight associated with each of the one or more first machine learning models; and

determining whether a content item is associated with a category based on the second machine learning model.

12. The system of claim 11, wherein each of the one or more first machine learning models is trained based on respective first training data including content items associated with the respective portion of the first period of time.

13. The system of claim 12, wherein the second machine learning model is trained based on second training data including content items associated with the second period of time.

14. The system of claim 13, wherein each of the one or more first machine learning models generates a score for a first content item in the second training data.

15. The system of claim 14, wherein the second training data indicates a score assigned to the first content item, and wherein the respective weight is determined based on a comparison of the score assigned to the first content item and the scores generated by the one or more first machine learning models for the first content item.

16. A non-transitory computer readable medium including instructions that, when executed by at least one hardware processor of a computing system, cause the computing system to perform a method comprising:

generating one or more first machine learning models, each of the one or more first machine learning models associated with a respective portion of a first period of time;

generating a second machine learning model incorporating the one or more first machine learning models as features, the second machine learning model associated with a second period of time;

determining a respective weight associated with each of the one or more first machine learning models; and

determining whether a content item is associated with a category based on the second machine learning model.

17. The non-transitory computer readable medium of claim 16, wherein each of the one or more first machine learning models is trained based on respective first training data including content items associated with the respective portion of the first period of time.

18. The non-transitory computer readable medium of claim 17, wherein the second machine learning model is trained based on second training data including content items associated with the second period of time.

19. The non-transitory computer readable medium of claim 18, wherein each of the one or more first machine learning models generates a score for a first content item in the second training data.

20. The non-transitory computer readable medium of claim 19, wherein the second training data indicates a score assigned to the first content item, and wherein the respective weight is determined based on a comparison of the score assigned to the first content item and the scores generated by the one or more first machine learning models for the first content item.