Problem solving method utilizing emergent clustering and case retrieval

• US 7,318,012 B2
• Filed: 01/21/2005
• Issued: 01/08/2008
• Est. Priority Date: 01/21/2005
• Status: Expired due to Fees

• Current Assignee: Techteam Government Solutions Incorporated (Jacobs Solutions, Inc.), Newvectors, LLC (Stefanini International Holdings Ltd. (Brazil))
• Original Assignee: Sven Brueckner
• Inventors: Brueckner, Sven
• Primary Examiner(s): THANGAVELU, KANDASAMY
• Application Number:
  US11/040,036
• Publication Number:
  US 20060167663A1
• Time in Patent Office: 1,082 Days
• Field of Search: 703/2, 703/11, 717/101, 715/530, 716/4, 438/200, 705/4, 705/52, 707/100, 707/102, 707/1, 707/3, 435/6, 713/176, 43/58
• US Class Current: 703/2
• CPC Class Codes: G06F 16/24578 using ranking
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: Y10S 707/99943 Generating database or data structure, e.g. via user interface

Abstract

An Adaptive, Any-Time Case Retrieval process combines existing knowledge (emergent clustering, case retrieval with CRN) in a novel and advantageous way. Although ideally suited to the Digital Body Development System (DBDS), the method is applicable to any problem solving environment involving iterative simulations. Given a description of a problem to be solved, a candidate solution is applied to generate a problem solving base case. The description of the problem is modified and the modification is recorded. Candidate solutions are applied to the problem as modified, thereby generating a problem solving event case. These steps are repeated so as to rank the relevance of the cases generated to arrive at an optimal solution to the problem. In the preferred embodiment a case retrieval network (CRN) structure is used in the ranking process.

Claims

1. A computer implemented method of selecting historical problem/solution cases for a design of an end product having a case base, comprising the steps of:

a) creating an arrangement of anchor points that combine observed deviations of the end product from functional specifications based on predefined measurement parameters of the design;

b) storing, in memory, a set of historical problem/solution cases associated with the design;

c) modifying the arrangement of the anchor points by a binning process and extracting a dynamic problem signature describing the current problem from the modified arrangement in bins;

d) retrieving the historical problem/solution cases and ranking the cases according to their similarity to the extracted signature, generating a rank score for each case, wherein the rank score is computed using data from the dynamic problem signature and problem signatures of historical cases;

e) repeating steps c) and d) until the rank score of a highest ranked case is equal to or greater than a pre-specified quality, where the quality is a function of selection probabilities of the cases in the case base;

f) selecting a problem/solution case that has a highest ranked score;

g) evaluating the selected problem/solution case based upon its divergence from the functional specifications;

h) repeating steps c) through g) until a specified number of problem/solution cases applicable to the design of the end product are selected; and

i) saving the selected problem/solution cases in a file.

2. The method of claim 1, wherein the anchor points include additional contextual data describing a problem domain.

3. The method of claim 1, wherein the step of selecting a problem/solution case is carried out probabilistically.

4. The method of claim 1, wherein the step of evaluating a selected case is based upon a manufacturing assembly simulation.

5. The method of claim 1, wherein the step of modifying the arrangement of the anchor points is based upon emergent clustering of anchor points.

6. The method of claim 5, wherein the anchor points are treated as clustering agents.

7. The method of claim 6, wherein the clustering agents are attracted to other, nearby agents.

8. The method of claim 7, wherein the dynamic problem signature is based upon locations of the clustering agents at a particular point in time.

9. The method of claim 8, wherein discrete binning of clustering agent locations produces the dynamic problem signature.

10. The method of claim 9, wherein the clustering agent locations are binned by a pheromone infrastructure (PI) and repeated deposits of digital pheromones at the anchor points.

11. The method of claim 10, wherein dynamic problem signature locations are recorded as a subset of locations of bins in the PI, wherein the dynamic problem signature locations determine the binning locations of the anchor points.

12. The method of claim 5, wherein:

case selection involves probabilities exhibiting a Shannon Entropy.

13. The method of claim 1, wherein each historical case has a recorded problem signature, wherein the recorded problem signature is based on the binning locations of anchor points of the historical case, wherein the anchor points of the historical case are based on deviations of the historical case end product from functional specifications.

14. The method of claim 1, wherein the historical cases are arranged in a case retrieval network (CRN).

15. The method of claim 14, wherein the CRN includes information, entity nodes representing observed locations of clusters of anchor points of the historical cases.

16. The method of claim 14, wherein the CRN is activated by the locations of the clusters of anchor points defined in the dynamic problem signature.

17. The method of claim 14, wherein the rank score of each historical case is determined by its current activation in the CRN where the current activation is a function of a distance between an observed location of a cluster of anchor points in the dynamic problem signature and location of a cluster of anchor points of the historical case in the recorded problem signature.

18. The method of claim 1, wherein:

The selection of a historical problem/solution case has a probability associated therewith; and

the probability is proportional to its rank score.

19. The method of claim 1, wherein the quality of the rank score is the Shannon Entropy of the selection probabilities associated with the cases in the case base.