Fuzzy logic entity behavior profiler

• US 5,577,169 A
• Filed: 04/29/1994
• Issued: 11/19/1996
• Est. Priority Date: 04/29/1994
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Prezioso, Robert T.
• Primary Examiner(s): Downs, Robert W.
• Assistant Examiner(s): KATBAB, ABDOLLAH
• Application Number:
  US08/236,731
• Time in Patent Office: 935 Days
• Field of Search: 395/61, 364/401, 364/406, 364/413.01
• US Class Current: 706/52
• CPC Class Codes: G06Q 10/10 Office automation; Time management

Abstract

The behavior of entities with common characteristics is profiled using fuzzy logic. First, behavior characteristics, or indicators of a target behavior, are determined. Norms and fuzzy sets are established for each behavior characteristic and then fuzzy logic is used to develop a series of scores. A behavior profile is developed from a collection of scores, and is organized either linearly or in a hierarchy, that can be used to compare the degree to which an entity within a peer group is, or is not, associated with a set of behavior characteristics relative to others in the same peer group. The number and organization of the behavior characteristics in the profile is expandable and the importance of characteristics can be weighted.

Claims

1. A method of profiling behavior of an entity comprising the steps of:

a. identifying a peer group having a plurality of entities, each entity exhibiting one or more behavioral characteristics;

b. determining a behavioral characteristic set of two or more of the behavioral characteristics that contribute to an entity profile that indicates a target behavior;

c. determining an amount of behavior characteristic each entity exhibits for each behavior characteristic in the behavior characteristic set;

d. defining a behavior characteristic fuzzy set for each behavior characteristic in the behavior characteristic set, each behavior characteristic fuzzy set defined using statistics of the respective amount of behavior characteristic exhibited by the entities in the peer group and each behavior characteristic fuzzy set being a function that represents an amount of membership one or more individual entities has in the respective behavior characteristic for an amount of behavior characteristic the individual entity exhibits;

e. using the behavior characteristic fuzzy sets to determine the amount of membership the individual entity has in each respective behavior characteristic in the behavior characteristic set;

f. scoring the amount of membership of each behavior characteristic for the individual entity to determine a score of each of the behavior characteristics in the behavior characteristic set for the individual entity, the score determined from a scoring fuzzy set having a scoring surface that maps the amount of membership of the behavior characteristic to the score; and

g. combining the the scores of each of the behavior characteristics in the behavior characteristic set for the individual entity to develop a composite used as an entity profile, the composite representing the degree of membership that the individual entity has in the target behavior.

2. A method, as in claim 1, where the composite is developed using fuzzy reasoning and defuzzification.

3. A method of profiling behavior of an entity comprising the steps of:

a. identifying a peer group having a plurality of entities, each entity exhibiting one or more behavioral characteristics;

b. determining a behavioral characteristic set of one or more of the behavioral characteristics that contribute to an entity profile that indicates a target behavior;

c. grouping the behavioral characteristics in the behavioral characteristic set into a hierarchy of one or more high level characteristics and one or more lower level characteristics, a lower level characteristic set being logically associated with a higher level characteristic and determining how much the associated higher level characteristic contributes to the profile;

d. determining the amount of each behavioral characteristic each entity exhibits for the behavioral characteristics in one or more lower level behavior characteristics;

e. defining a behavior characteristic fuzzy set for each behavior characteristic in the lower level characteristic sets that represents the amount of membership an entity has in the respective behavioral characteristic, each behavior characteristic fuzzy set defined using statistics of the respective amount of behavior characteristic exhibited by the entities in the peer group and each behavior characteristic fuzzy set being a function that represents an amount of membership that one or more individual entities has in the respective behavior characteristic for an amount of behavior characteristic the individual entity exhibits;

f. using the behavior characteristic fuzzy sets to determine the amount of membership the individual entity has in each respective behavior characteristic in the lower level characteristic set;

g. using a scoring fuzzy set to determine a score that an entity has for each respective behavior characteristic in the lower level characteristic set based on the respective amount of membership of the entity has in the behavior characteristic, the scoring fuzzy set having a scoring surface that maps the amount of membership of the behavior characteristic to the score;

h. developing, from the scores of the lower level characteristics in the hierarchy, a group score for each associated high level characteristic; and

i. establishing a profile of an entity'"'"'s behavior using a composite score of the group scores in the hierarchy by using fuzzy reasoning and defuzzification.

4. A method, as in claim 3, where the score of one or more behavior characteristics of an entity is weighted.

5. A method, as in claim 4, where the weighting is done by hedging the scoring fuzzy set.

6. A method, as in claim 5, where the hedging is linguistic hedging.

7. A method, as in claim 3, where the hierarchy is expandable.

8. A method, as in claim 7, where the hierarchy is expandable by adding records that contain information defining a lower level behavior characteristic and the group to which the lower level behavior characteristic belongs.

9. A method, as in claim 3, where the fuzzy set has any one of the types of fuzzy set surfaces including linear, inclining growth, declining growth, bell, and inverted bell.

10. A method, as in claim 9, where the type of fuzzy set is user selected.

11. A method, as in claim 3, where the fuzzy set is defined by a set of truth pairs.

12. A method, as in claim 5, where the hedging is any one of the hedge types including VERY, SOMEWHAT, EXTREMELY, SLIGHTLY, and NOT.

13. A method, as in claim 12, where the hedge type is user selected.

14. A method, as in claim 3, where the composite score of the group scores is determined using hedged group score fuzzy sets.

15. A method, as in claim 3, where the hierarchy has a plurality of levels.

16. A system for profiling the target behavior of one or more entities comprising:

a. a computer having a central processing unit, a memory, and a user interface;

b. a plurality of entity data structures, stored in memory, each defining an entity in a plurality of entities in a peer group;

c. a plurality of fuzzy set data structures, stored in memory, each defining a degree of entity membership in a behavior characteristic for a given amount of the behavior characteristic that the entity exhibits, the behavior characteristic being one of a plurality of behavior characteristics contributing to the profile of the target behavior of the entity;

d. a plurality of hierarchy data structures, stored in memory, each defining a group of behavior characteristics to which the behavior characteristic belongs in a hierarchy of behavior characteristics;

e. a plurality of behavior characteristic amount data structures, stored in memory, each defining the amount of a behavior characteristic exhibited by an entity;

f. an algorithm, executed by the central processing unit to perform the following steps;

1. using the hierarchy data structures to group the behavioral characteristics into a hierarchy of one or more high level characteristics and one or more lower level characteristics, a lower level characteristic set being logically associated with a higher level characteristic and determining how much the associated higher level characteristic contributes to the profile;

2. using the amount data structures to determine the amount of each behavior characteristic each entity exhibits for the behavior characteristics in one or more lower level behavior characteristics, the entities determined from the entity data structures;

18. A system, as in claim 16, where one or more entity profiles is displayed to the user on the user interface.

22. A system, as in claim 16, where the profiled entities are physicians and the behavior characteristics are physician behavior.

24. A system, as in claim 16, where the profiled entities are employees and the behavior characteristics are employee behavior.

28. A system, as in claim 16, where the profiled entities are business units and the behavior characteristics indicate performance.

30. A system, as in claim 16, where the profiled entities are securities and the behavior characteristics indicate security portfolio performance.

32. A method, as in claim 30, where the profile of one or more entities is trended over time.

33. A method, as in claim 30, where the profiles of a plurality of entities is ranked in a list.

17. using the fuzzy set data structures to define a fuzzy set for each behavior characteristic in the lower level behavior characteristic sets that represents the amount of membership an entity has in the respective behavior characteristic;

4. using the fuzzy sets to determine the amount of membership one or more entities has in each respective behavior characteristic in the lower level characteristic set;

5. using a scoring fuzzy set to determine a score that an entity has for each respective behavior characteristic in the lower level characteristic set based on the respective amount of membership of the entity has in the behavior characteristic;

6. developing, from the scores of the lower level characteristics in the hierarchy, a group score for each associated high level characteristic; and

7. establishing a profile of an entity'"'"'s target behavior with a composite score of the group scores in the hierarchy determined by using fuzzy reasoning and defuzzification.

19. A system, as in claim 17, where the user interface is a graphical display.

20. A system, as in claim 17, where the user interface is a printer.

21. A system, as in claim 17, where one or more entity profiles are displayed to the user in an order ranked by degree of target behavior.

23. A system, as in claim 21, where the physician behavior indicates fraud.

25. A system, as in claim 23, where the employees are cashiers and the behavior characteristics indicate fraud.

26. A system, as in claim 23, where the employees are purchasing agents and the behavior characteristics indicate purchasing transactions.

27. A system, as in claim 23, where the employees are sales people and the behavior characteristics indicate sales activity.

29. A system, as in claim 27, where the business units can be any one of the set of:

departments in a corporation, clinics, hospitals, medical supply companies, ambulance service companies, laboratories, nursing homes, and pharmacies.

31. A method of using an entity behavior profiler comprising the steps of:

a. identifying a peer group having a plurality of entities, each entity exhibiting one or more behavioral characteristics;

b. determining a behavioral characteristic set of one or more behavioral characteristics that contribute to an entity profile that indicates a target behavior;

c. grouping the behavioral characteristics in the behavioral characteristic set into a hierarchy of one or more high level characteristics and one or more lower level characteristics, a lower level characteristic set being logically associated with a higher level characteristic and determining how much the associated higher level characteristic contributes to the profile;

d. determining the amount of each behavior characteristic each entity exhibits for the behavior characteristics in one or more lower level behavior characteristics;

e. defining a behavior characteristic fuzzy set for each behavior characteristic in the lower level behavior characteristic sets that represents the amount of membership an entity has in the respective behavior characteristic, each behavior characteristic fuzzy set defined using statistics of the respective amount of behavior characteristic exhibited by the entities in the peer group and each behavior characteristic fuzzy set being a function that represents an amount of membership that one or more individual entities has in the respective behavior characteristic for an amount of behavior characteristic the individual entity exhibits;

f. using the behavior characteristic fuzzy sets to determine the amount of membership the individual entity has in each respective behavior characteristic in the lower level characteristic set;

g. using a scoring fuzzy set to determine a score that an entity has for each respective behavior characteristic in the lower level characteristic set based on the respective amount of membership of the entity has in the behavior characteristic, the scoring fuzzy set having a scoring surface that maps the amount of membership of the behavior characteristic to the score;

h. developing, from the scores of the lower level characteristics in the hierarchy, a group score for each associated high level characteristic;

i. establishing a profile of an entity'"'"'s behavior using a composite score of the group scores in the hierarchy by using fuzzy reasoning and defuzzification; and

j. selecting a subset of entities in the peer group that exhibit a level of target behavior indicated by the profile.