System and method for real time pattern identification

• US 7,930,260 B2
• Filed: 02/14/2008
• Issued: 04/19/2011
• Est. Priority Date: 02/14/2008
• Status: Active Grant

• Current Assignee: eBay Inc.
• Original Assignee: eBay Inc.
• Inventors: Deo, Harshal Ulhas, Carnes, Brian Matthew, Maddali, Lakshman Shyam Sundar
• Primary Examiner(s): HOLMES, MICHAEL B
• Application Number:
  US12/031,258
• Publication Number:
  US 20090210368A1
• Time in Patent Office: 1,160 Days
• Field of Search: 706/20
• US Class Current: 706/20
• CPC Class Codes: G06F 18/10 Pre-processing; Data cleansing
• CPC Class Codes: G06F 18/24 Classification techniques
• CPC Class Codes: G06N 3/08 Learning methods
• CPC Class Codes: G06N 5/046 Forward inferencing; Production systems

Abstract

A method for near real time patterns identification, in one example embodiment, comprises receiving a data stream containing information associated with a transaction and participants of the transaction and receiving an Artificial Intelligence (AI) algorithm trained to score data in the data stream. The method may further comprise receiving metadata associated with the historical information, comparing the data stream to the metadata by measuring differences between variables included in the historical metadata and the data stream. The method may further comprise modifying the data stream to suit the AI algorithm when the differences between variables are below predetermined threshold values and retraining the AI algorithm based on the data stream when the differences between the variables are greater than the predetermined threshold values. The method may further comprise feeding the data stream to the AI algorithm to classify the variables in the data stream.

Claims

1. A computer-implemented method comprising:

receiving a first data stream containing information associated with a transaction and participants of the transaction;

receiving an Artificial Intelligence (AI) algorithm, the AI algorithm being trained to classify the first data stream based on historical information related to the transaction and the participants of the transaction;

receiving metadata associated with the historical information utilized in training of the AI algorithm;

comparing the first data stream to the metadata associated with the historical information; and

based on the comparison, selectively modifying the first data stream to suit the AI algorithm, thereby producing a modified data stream.

2. The computer-implemented method of claim 1, wherein the comparison comprises:

generating a first metadata based on the first data stream;

measuring differences between first variables included in the first metadata and second variables included in the metadata associated with the historical information; and

comparing the differences between the first variables and the second variables to predetermined threshold values.

3. The computer-implemented method of claim 2, further comprising:

modifying the first data stream to suit the AI algorithm when the differences between the first variables and the second variables are below the predetermined threshold values, thereby producing the modified data stream; and

classifying the first variables in the modified data stream by the AI algorithm.

4. The computer-implemented method of claim 2, further comprising:

retraining the AI algorithm based on the first data stream when the differences between the first variables and the second variables are greater than the predetermined threshold values; and

classifying the first variables in the first data stream by the AI algorithm.

5. The computer-implemented method of claim 1, wherein the modifying includes adjusting averages and standard distributions of the first variables to correspond to the averages and standard distributions of the second variables.

6. The computer-implemented method of claim 1, further comprising automatically converting the first data stream and the historical information into a canonical form.

7. The computer-implemented method of claim 3, further comprising:

determining behavioral patterns based on the variables classified by the Al algorithm; and

deploying the behavioural patterns to a transaction server in a continuous data stream to be utilized in business rules of the transaction server.

8. The computer-implemented method of claim 4, further comprising:

determining behavioral patterns based on the variables classified by the Al algorithm; and

deploying the behavioural patterns to a transaction server in a continuous data stream to be utilized in business rules of the transaction server.

9. The computer-implemented method of claim 1, further comprising saving the first data stream for a predetermined time to delay classifying of the transaction until transaction feedback is received.

10. The computer-implemented method of claim 9, further comprising retraining the AI algorithm based on the feedback.

11. A computer-implemented system comprising:

a data stream receiving module to receive

a first data stream containing information associated with a transaction and participants of the transaction,

an Artificial Intelligence (AI) algorithm, the AI algorithm being trained to classify the first data stream based on historical information related to the transaction and the participants of the transaction, and

metadata associated with the historical information utilized in training of the AI algorithm;

an evaluation module to compare the first data stream to the metadata associated with the historical information; and

a modifying module to selectively modify the first data stream to suit the AI algorithm based on comparison, thereby producing a modified data stream.

12. The computer-implemented system of claim 11, wherein the comparing module is to generate a first metadata based on the first data stream, to measure differences between first variables included in the first metadata and second variables included in the metadata associated with the historical information, and to compare the differences between the first variables and the second variables to predetermined threshold values.

13. The computer-implemented system of claim 12, wherein the modifying module is to modify the first data stream to suit the AI algorithm when the differences between the first variables and the second variables are below the predetermined threshold values, thereby producing the modified data stream;

and a classification module to classify the first variables in the modified data stream by the AI algorithm.

14. The computer-implemented system of claim 12, further comprising:

a training module to retrain the AI algorithm based on the first data stream when the differences between the first variables and the second variables are greater than the predetermined threshold values; and

a classification module to classify the first variables in the first data stream by the AI algorithm.

15. The computer-implemented system of claim 11, wherein the modifying module is to adjust averages and standard distributions of the first variables to correspond to the averages and standard distributions of the second variables.

16. The computer-implemented system of claim 11, further comprising a converting module to automatically convert the first data stream and the historical information into a canonical form.

17. The computer-implemented system of claim 13, further comprising:

the classification module to determine behavioral patterns based on the variables classified by the AI algorithm; and

a deployment module to deploy the behavioural patterns to a transaction server in a continuous data stream to be utilized in business rules of the transaction server.

18. The computer-implemented system of claim 14, further comprising:

the classifying module to determine behavioral patterns based on the variables classified by the AI algorithm; and

the deployment module to deploy the behavioural patterns to a transaction server in a continuous data stream to be utilized in business rules of the transaction server.

19. The computer-implemented system of claim 11, further comprising a saving module to save the first data stream for a predetermined time to delay classifying of the transaction until transaction feedback is received.

20. The computer-implemented system of claim 19, further comprising a training module to retrain the AI algorithm based on the feedback.

21. A machine-readable medium comprising instructions, which when implemented by one or more processors, perform the following operations:

receive a first data stream containing information associated with a transaction and participants of the transaction;

receive an Artificial Intelligence (Al) algorithm, the AI algorithm being trained to classify the first data stream based on historical information related to the transaction and the participants of the transaction;

receive metadata associated with the historical information utilized in training of the AI algorithm;

compare the first data stream to the metadata associated with the historical information; and

based on the comparison, selectively modify the first data stream to suit the AI algorithm, thereby producing a modified data stream.

22. An apparatus comprising:

means for receiving a first data stream containing information associated with a transaction and participants of the transaction;

means for receiving an Artificial Intelligence (Al) algorithm, the AI algorithm being trained to classify the first data stream based on historical information related to the transaction and the participants of the transaction;

means for receiving metadata associated with the historical information utilized in training of the AI algorithm;

means for comparing the first data stream to the metadata associated with the historical information; and

means for selectively modifying the first data stream to suit the AI algorithm based on the comparison, thereby producing a modified data stream.