Method and apparatus for generating a bi-gram score in fraud risk analysis

• US 8,244,629 B2
• Filed: 06/24/2011
• Issued: 08/14/2012
• Est. Priority Date: 07/28/1997
• Status: Expired due to Fees

• Current Assignee: Frank King, John Pettit, Michael Lewis, Thomas Arnold, William Wright
• Original Assignee: Frank King, John Pettit, Michael Lewis, Thomas Arnold, William Wright
• Inventors: Lewis, Michael, King, Frank, Arnold, Thomas, Wright, William, Pettit, John
• Primary Examiner(s): Patel, Jagdish
• Application Number:
  US13/168,795
• Publication Number:
  US 20110276468A1
• Time in Patent Office: 417 Days
• Field of Search: 705/26.1, 705/38, 704/236
• US Class Current: 705/38
• CPC Class Codes: G06Q 10/0635 Risk analysis of enterprise or organisation activities
• CPC Class Codes: G06Q 20/10 specially adapted for electronic funds transfer [EFT] systems; specially adapted for home banking systems
• CPC Class Codes: G06Q 20/12 specially adapted for electronic shopping systems
• CPC Class Codes: G06Q 20/3674 involving authentication
• CPC Class Codes: G06Q 20/3829 involving key management
• CPC Class Codes: G06Q 20/40 Authorisation, e.g. identification of payer or payee, verification of customer or shop credentials; Review and approval of payers, e.g. check credit lines or negative lists
• CPC Class Codes: G06Q 20/4016 involving fraud or risk level assessment in transaction processing
• CPC Class Codes: G06Q 30/0185 Product, service or business identity fraud
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 40/00 Finance; Insurance; Tax strategies; Processing of corporate or income taxes
• CPC Class Codes: G06Q 40/03 Credit; Loans; Processing thereof

Abstract

Evaluating fraud risk in a transaction between consumer and a merchant over a network is disclosed. The merchant requests service over the network using a secure, open messaging protocol. An e-commerce transaction or electronic purchase order is received from the merchant, the level of risk associated with each order is measured, and a risk score is returned to the merchant. In one embodiment, data validation, highly predictive artificial intelligence pattern matching, network data aggregation and negative file checks are used to examine numerous factors to calculate fraud risk. A risk score is generated and compared to the merchant'"'"'s specified risk threshold. The result is returned to the merchant for order disposition.

Claims

1. A method, comprising:

receiving, using one or more processing units, character data;

determining, using the one or more processing units, a set of character pairs, wherein each character pair in the set represents a pair of adjacent characters in the character data;

generating, using the one or more processing units, bi-gram probability data using data from one or more data sources, wherein the data from the one or more data sources includes one or more words, and wherein the bi-gram probability data includes a plurality of bi-grams, wherein each bi-gram is a pair of letters that is associated with one or more probability values;

matching, using the one or more processing units, each character pair to a bi-gram in the bi-gram probability data;

determining, using the one or more processing units, a probability value for each character pair using the bi-gram that matches that character pair, wherein a probability value represents occurrences of a character pair in the one or more data sources, and wherein a probability value is used to determine a measure of intelligibility associated with the character data; and

generating, using the one or more processing units, a bi-gram score using the determined probability values, wherein the bi-gram score represents a measure of intelligibility associated with the character data.

2. The method of claim 1, wherein when a character pair does not match any bi-grams in the bi-gram probability data, the letter pair has a pre-determined probability value.

3. The method of claim 1, wherein each probability value associated with a bi-gram corresponds to one or more occurrences of the bi-gram at a particular location within a word.

4. The method of claim 1, further comprising:

determining, using the one or more processing units, whether the character data is an acronym; and

setting, using the one or more processing units, the bi-gram score to a pre-determined default score when an acronym is detected.

5. The method of claim 1, further comprising:

determining, using the one or more processing units, whether the character data has an ethnic association, wherein when the character data has an ethnic association, the bi-gram probability data is generated using an ethnic data source.

6. The method of claim 1, wherein the character data has a data type, wherein the data type has an associated data type source, and wherein the bi-gram probability data is generated using the data type source.

7. The method of claim 1, further comprising:

receiving, using the one or more processing units, information data, wherein the information data includes the character data;

determining, using the one or more processing units, a historical score based upon a comparison of the information data to historical information;

generating, using the one or more processing units, a statistical estimate by applying a statistical model to the bi-gram score and the historical score;

generating, using the one or more processing units, a heuristic estimate by applying a heuristic model to the bi-gram score and the historical score;

defining, using the one or more processing units, a plurality of fraud risk zones, wherein each fraud risk zone has a corresponding blending policy, and wherein a blending policy dictates a magnitude and an allowable direction of influence for a statistical estimate and a heuristic estimate;

determining, using the one or more processing units, a fraud risk zone and a corresponding blending policy for the heuristic estimate; and

generating, using the one or more processing units, a final score by applying the blending policy to the statistical estimate and the heuristic estimate.

8. The method of claim 7, wherein a blending policy is defined using a characteristic of the corresponding fraud risk zone, a characteristic of the statistical model, and a characteristic of the heuristic model.

9. The method of claim 7, wherein each fraud risk zone represents a likelihood of fraudulent activity.

10. The method of claim 9, wherein fraud risk zones having higher likelihoods of fraudulent activity have corresponding blending policies that give greater weight to a heuristic estimate, relative to a statistical estimate;

and wherein fraud risk zones having lower likelihoods of fraudulent activity have corresponding blending policies that give less weight to a heuristic estimate, relative to a statistical estimate.

11. A system, comprising:

one or more processors;

a non-transitory computer-readable storage medium containing instructions configured to cause the one or more processors to perform operations, including;

receiving character data;

determining a set of character pairs, wherein each character pair in the set represents a pair of adjacent characters in the character data;

generating bi-gram probability data using data from one or more data sources, wherein the data from the one or more data sources includes one or more words, and wherein the bi-gram probability data includes a plurality of bi-grams, wherein each bi-gram is a pair of letters that is associated with one or more probability values;

matching each character pair to a bi-gram in the bi-gram probability data;

determining a probability value for each character pair using the bi-gram that matches that character pair, wherein a probability value represents occurrences of a character pair in the one or more data sources, and wherein a probability value is used to determine a measure of intelligibility associated with the character data; and

generating a bi-gram score using the determined probability values, wherein the bi-gram score represents a measure of intelligibility associated with the character data.

12. The system of claim 11, wherein when a character pair does not match any bi-grams in the bi-gram probability data, the letter pair has a pre-determined probability value.

13. The system of claim 11, wherein each probability value associated with a bi-gram corresponds to one or more occurrences of the bi-gram at a particular location within a word.

14. The system of claim 11, further comprising instructions to cause the one or more processors to perform operations, including:

determining whether the character data is an acronym; and

setting the bi-gram score to a pre-determined default score when an acronym is detected.

15. The system of claim 11, further comprising instructions to cause the one or more processors to perform operations, including:

determining whether the character data has an ethnic association, wherein when the character data has an ethnic association, the bi-gram probability data is generated using an ethnic data source.

16. The system of claim 11, wherein the character data has a data type, wherein the data type has an associated data type source, and wherein the bi-gram probability data is generated using the data type source.

17. The system of claim 11, further comprising instructions to cause the one or more processors to perform operations, including:

receiving information data, wherein the information data includes the character data;

determining a historical score based upon a comparison of the information data to historical information;

generating a statistical estimate by applying a statistical model to the bi-gram score and the historical score;

generating a heuristic estimate by applying a heuristic model to the bi-gram score and the historical score;

defining a plurality of fraud risk zones, wherein each fraud risk zone has a corresponding blending policy, and wherein a blending policy dictates a magnitude and an allowable direction of influence for a statistical estimate and a heuristic estimate;

determining a fraud risk zone and a corresponding blending policy for the heuristic estimate; and

generating a final score by applying the blending policy to the statistical estimate and the heuristic estimate.

18. The system of claim 17, wherein a blending policy is defined using a characteristic of the corresponding fraud risk zone, a characteristic of the statistical model, and a characteristic of the heuristic model.

19. The system of claim 17, wherein each fraud risk zone represents a likelihood of fraudulent activity.

20. The system of claim 19, wherein fraud risk zones having higher likelihoods of fraudulent activity have corresponding blending policies that give greater weight to a heuristic estimate, relative to a statistical estimate;

and wherein fraud risk zones having lower likelihoods of fraudulent activity have corresponding blending policies that give less weight to a heuristic estimate, relative to a statistical estimate.

21. A computer program product, tangibly embodied in a non-transitory machine readable storage medium, including instructions configured to cause a data processing apparatus to:

receive character data;

determine a set of character pairs, wherein each character pair in the set represents a pair of adjacent characters in the character data;

generate bi-gram probability data using data from one or more data sources, wherein the data from the one or more data sources includes one or more words, wherein the bi-gram probability data includes a plurality of bi-grams, wherein each bi-gram is a pair of letters that is associated with one or more probability values;

match each character pair to a bi-gram in the bi-gram probability data;

determine a probability value for each character pair using the bi-gram that matches that character pair, wherein a probability value represents occurrences of a character pair in the one or more data sources, and wherein a probability value is used to determine a measure of intelligibility associated with the character data; and

generate a bi-gram score using the determined probability values, wherein the bi-gram score represents a measure of intelligibility associated with the character data.

22. The computer program product of claim 21, wherein when a character pair does not match any bi-grams in the bi-gram probability data, the letter pair has a pre-determined probability value.

23. The computer program product of claim 21, and wherein each probability value associated with a bi-gram corresponds to one or more occurrences of the bi-gram at a particular location within a word.

24. The computer program product of claim 21, further comprising instructions configured to cause a data processing apparatus to:

determine whether the character data is an acronym; and

set the bi-gram score to a pre-determined default score when an acronym is detected.

25. The computer program product of claim 21, further comprising instructions configured to cause a data processing apparatus to:

determine whether the character data has an ethnic association, wherein when the character data has an ethnic association, the bi-gram probability data is generated using an ethnic data source.

26. The computer program product of claim 21, wherein the character data has a data type, wherein the data type has an associated data type source, and wherein the bi-gram probability data is generated using the data type source.

27. The computer program product of claim 21, further comprising instructions configured to cause a data processing apparatus to:

receive information data, wherein the information data includes the character data;

determine a historical score based upon a comparison of the information data to historical information;

generate a statistical estimate by applying a statistical model to the bi-gram score and the historical score;

generate a heuristic estimate by applying a heuristic model to the bi-gram score and the historical score;

define a plurality of fraud risk zones, wherein each fraud risk zone has a corresponding blending policy, and wherein a blending policy dictates a magnitude and an allowable direction of influence for a statistical estimate and a heuristic estimate;

determine a fraud risk zone and a corresponding blending policy for the heuristic estimate; and

generating a final score by applying the blending policy to the statistical estimate and the heuristic estimate.

28. The computer program product of claim 27, wherein a blending policy is defined using a characteristic of the corresponding fraud risk zone, a characteristic of the statistical model, and a characteristic of the heuristic model.

29. The computer program product of claim 27, wherein each fraud risk zone represents a likelihood of fraudulent activity.

30. The computer program product of claim 29, wherein fraud risk zones having higher likelihoods of fraudulent activity have corresponding blending policies that give greater weight to a heuristic estimate, relative to a statistical estimate;

and wherein fraud risk zones having lower likelihoods of fraudulent activity have corresponding blending policies that give less weight to a heuristic estimate, relative to a statistical estimate.