System and method for pattern detection in user accounts through fuzzy pattern matching

Patent metadata

Field	Value
Publication number	US-11799975-B1
Application number	US-202217662155-A
Country	US
Kind code	B1
Filing date	May 5, 2022
Priority date	May 5, 2022
Publication date	Oct 24, 2023
Grant date	Oct 24, 2023

Abstract

Various techniques are disclosed for providing dynamic, real-time pattern detection and linking between newly created user accounts and existing user accounts. Certain solutions include assessing matches of a new user account to nodes in a graphical representation of a machine learning algorithm based on predefined patterns in the properties of the new user account. The nodes in the graphical representation may each have different predefined patterns that have been determined based on patterns in previous information for user accounts, which can also be updated in real-time as needed. Accordingly, when a new user account is matched (e.g., assigned) to a node that has accounts with known issues associated with the node, the new user account may be flagged for increased scrutiny or other solutions.

First claim

What is claimed is:

1. A method, comprising:
receiving, by a computer system, account information corresponding to a new user account generated for a service system, wherein the account information includes values for a plurality of account properties;
determining whether the new user account belongs to a node in a graphical detection database, wherein the node is associated with a predefined pattern in two or more of the account properties, wherein the predefined pattern includes a combination of characters from the two or more account properties, the node holding existing values for the two or more account properties based on the predefined pattern, wherein the existing values held by the node for the two or more account properties based on the predefined pattern include existing values for the combination of characters, and wherein determining whether the new user account belongs to the node includes:
extracting values for the two or more account properties from the new user account based on the predefined pattern, wherein the extracted values for the two or more account properties from the new user account based on the predefined pattern include extracted values for the combination of characters; and
determining the new user account belongs to the node when the extracted values of the two or more account properties corresponding to the new user account match the existing values for the two or more account properties held in the node, wherein the extracted values of the two or more account properties corresponding to the new user account match the existing values for the two or more account properties held in the node when the extracted values for the combination of characters match the existing values for the combination of characters; and

assigning an account issue determination to the new user account based on additional user account information associated with the node in response to the new user account being determined to belong to the node.

2. The method of claim 1 , wherein the existing values for the two or more account properties based on the predefined pattern include a first existing value for a first account property of the two or more of the account properties and a second existing value for a second account property of the two or more of the account properties.

3. The method of claim 1 , further comprising assigning the new user account to an account node in the graphical detection database that branches from the node in response to the new user account being determined to belong to the node.

4. The method of claim 3 , wherein the additional user account information is associated with additional account nodes in the graphical detection database that branch from the node.

5. The method of claim 1 , wherein the graphical detection database includes at least one additional node, the at least one additional node is associated with an additional predefined pattern in two or more of the account properties, the method further comprising determining whether the new user account belongs to the at least one additional node.

6. The method of claim 5 , wherein determining whether the new user account belongs to the at least one additional node includes:
extracting values for the two or more account properties from the new user account based on the additional predefined pattern; and
determining the new user account belongs to the at least one additional node when the extracted values of the two or more account properties corresponding to the new user account match the existing values for the two or more account properties held in the at least one additional node.

7. The method of claim 1 , wherein the account issue determination assigned to the new user account is an increased risk for fraud when the additional user account information associated with the node includes information indicative of known fraud.

8. The method of claim 1 , further comprising providing the account information for the new user account to a solution service configured to resolve the account issue determination assigned to the new user account, wherein the solution service is one of a fraud risk solution service, a collusion detection service, or a dispute resolution service.

9. A non-transitory computer-readable medium having instructions stored thereon that are executable by a computing device to perform operations, comprising:
receiving account information corresponding to a new user account generated for a service system, wherein the account information includes values for a plurality of account properties;
determining whether the new user account belongs to one or more pattern nodes in a graphical detection database, wherein the pattern nodes in the graphical detection database are associated with predefined patterns in the account properties, wherein the predefined patterns include combinations of characters from the account properties, and wherein the pattern nodes hold existing values for the account properties based on the predefined patterns, wherein the existing values held by the pattern nodes for the account properties based on the predefined patterns include existing values for the combinations of characters, and wherein determining whether the new user account belongs to a specified pattern node includes:
extracting values for the account properties from the new user account based on a predefined pattern for the specified pattern node, wherein the extracted values for the account properties from the new user account based on the predefined pattern for the specified pattern node include extracted values for a combination of characters for the specified pattern node; and
determining the new user account belongs to the specified pattern node when the extracted values of the account properties corresponding to the new user account match the existing values for the properties held in the specified pattern node, wherein the extracted values of the account properties corresponding to the new user account match the existing values for the account properties held in the specified pattern node when the extracted values for the combination of characters match the existing values for the combination of characters in the specified pattern node; and

assigning the new user account to an account node linked to the specified pattern node in response to the new user account being determined to belong to the specified pattern node.

10. The non-transitory computer-readable medium of claim 9 , wherein at least two pattern nodes in the graphical detection database have different predefined patterns.

11. The non-transitory computer-readable medium of claim 9 , wherein a first pattern node holds existing values for a first set of two or more account properties in a first predefined pattern associated with the first pattern node, and wherein a second pattern node holds existing values for a second set of two or more account properties in a second predefined pattern associated with the second pattern node.

12. The non-transitory computer-readable medium of claim 9 , wherein a first pattern node holds first existing values for two or more account properties based on a predefined pattern, and wherein a second pattern node holds second existing values for the two or more account properties based on the predefined pattern, the first values being different than the second values.

13. The non-transitory computer-readable medium of claim 9 , wherein information for additional user accounts associated with additional account nodes linked to the specified pattern node includes indicators for known fraud by the additional user accounts, the operations further comprising assigning an increased fraud ris

Assignees

• Paypal Inc

Inventors

• Lin Ying
• Zhang Jiaqi

Classifications

• H04L67/306 Primary

  User profiles · CPC title

• G06F18/29

  Graphical models, e.g. Bayesian networks · CPC title

• G06N5/048

  Fuzzy inferencing · CPC title

• G06N20/00

  Machine learning · CPC title

• G06N7/02

  using fuzzy logic (computing arrangements based on biological models G06N3/00; computing arrangements using knowledge-based models G06N5/00) · CPC title