Method and tool for data mining in automatic decision making systems

• US 6,820,070 B2
• Filed: 12/08/2000
• Issued: 11/16/2004
• Est. Priority Date: 06/07/2000
• Status: Expired due to Fees

• Current Assignee: ADA ANALYTICS ISRAEL LTD (Advanced Emissions Solutions, Inc.)
• Original Assignee: Insyst, Ltd. (Advanced Emissions Solutions, Inc.)
• Inventors: Goldman, Arnold J., Hartman, Jehuda, Fisher, Joseph, Sarel, Shlomo
• Primary Examiner(s): Davis, George B.
• Application Number:
  US09/731,978
• Publication Number:
  US 20010054032A1
• Time in Patent Office: 1,439 Days
• Field of Search: 706/46
• US Class Current: 706/46
• CPC Class Codes: G05B 15/02 electric
• CPC Class Codes: G05B 19/00 Programme-control systems
• CPC Class Codes: G05B 19/41885 characterised by modeling, simulation of the manufacturing system
• CPC Class Codes: G05B 2219/31338 Design, flexible manufacturing cell design
• CPC Class Codes: G05B 2219/31339 From parameters, build processes, select control elements and their connection
• CPC Class Codes: G05B 2219/31353 Expert system to design cellular manufacturing systems
• CPC Class Codes: G05B 2219/32345 Of interconnection of cells, subsystems, distributed simulation
• CPC Class Codes: G05B 2219/33027 Artificial neural network controller
• CPC Class Codes: G05B 2219/33079 Table with functional, weighting coefficients, function
• CPC Class Codes: G05B 2219/45232 CMP chemical mechanical polishing of wafer
• CPC Class Codes: Y02P 90/02 Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]

Abstract

In an automatic decision-making system, a method and a tool for the reduction of the dimension of data mining, which is automatically coupled to an empirical predictor of the system. The method includes a qualitative modeling of the interrelations between various objects whose attributes are relevant to a score made by the predictor according to which decisions are made, wherein this relevancy is determined by an input of a domain expert to the problem in hand. The model is called a Knowledge-Tree and its conclusions are represented by a graphical symbolization called the Knowledge-Tree map. Data mining, which follows the construction of the Knowledge-Tree map regards only datasets which are associated with logical and validated branches of the knowledge tree. Because the expert input which reduces the dimension of data mining was completed prior to data mining, interception by human reasoning is not needed after data mining and the decision making process can proceed automatically.

Claims

1. A method for automated decision-making comprising the steps of:

constructing a series of interconnection cells and using values obtained thereby as decision makers, said construction comprising;

a) modeling of relations between a plurality of objects in a system, each object having at least one outcome, each object being subjected to at least one influential factor affecting said at least one outcome and building said modeled relations as interconnections between said interconnection cells;

b) carrying out computerized data mining in datasets associated with said modeled relations between said at least one outcome and said at least one influential factor of at least one said object, and constraining said data mining to relations modeled by said interconnections;

c) building a quantitative model to predict a score for said at least one outcome, by attributing values from said data mining to said relations; and

d) making an output decision for said system according to said score of said at least one outcome of said at least one object.

2. The method as in claim 1 wherein said modeling of relations comprises:

(i) selecting at least two of said plurality of objects;

(ii) for each of said at least two objects, defining at least one outcome;

(iii) for each of said at least one outcome, identifying at least one influential factor;

(iv) validating an influence of said at least one possible influential factor on each of said at least one outcome respectively; and

,

(v) symbolizing graphically said at least two objects, said outcomes of said at least two objects and said influences of said outcomes of said at least two objects.

3. The method as in claim 2 wherein said selecting of said plurality of objects with in said system is based on knowledge selected from the group consisting of disciplinary knowledge and structural knowledge that are appropriate for a specific functional operation of said system.

4. The method as in claim 2 wherein said disciplinary knowledge is selected from the group consisting of warehouse data and expert experience.

5. The method as in claim 2 wherein said structural knowledge is selected from the group consisting of functional, configurational, logical and heuristic structure.

6. The method as in claim 2 wherein said at least one outcome of a said at least one object is defined by an expert having expertise in a domain of said at least one object.

7. The method as in claim 2 wherein said at least one influential factor on said at least one outcome of said at least one object is defined by an expert having expertise in a domain of said at least one object.

8. The method as in claim 2 wherein said validating of an influence of said at least one possible influential factor on said at least one outcome includes seeking for a correlation between said at least one possible influential factor and said at least one outcome.

9. The method as in claim 2 comprising constructing said interconnection cells such that one of said at least one outcomes of a first of said plurality of objects is an influence on one of said at least one outcomes of a second of said plurality of objects.

10. The method as in claim 2 wherein said graphical symbolization is stored in a memory of a computer.

11. The method as in claim 1 wherein said data mining is effected using statistical techniques selected from the group consisting of linear regression, nearest neighbor, clustering, process output empirical modeling (POEM), classification and regression tree (CART), chi-square automatic interaction detector (CHAID), decision trees and neural network empirical modeling.

12. The method as in claim 1 wherein said building of said quantitative model is effected using statistical techniques selected from the group consisting of linear regression, nearest neighbor, clustering, process output empirical modeling (POEM), classification and regression tree (CART), chi-square automatic interaction detector (CHAID), decision trees and neural network empirical modeling.

13. A knowledge engineering tool for defining a relationship pattern among a plurality of objects, said tool comprising:

a graphical symbolization unit for symbolizing said objects and assumed interactions thereof by constructing at least two interconnection cells to represent components of a system whose relationship pattern is to be defined and connections therebetween to indicate said assumed interactions;

and a numerical processor, associated with said graphical symbolization unit for carrying out computerized data mining amongst data representative of said system, and using said relationship pattern as a dimension reduction constraint on said data mining, said computerized data mining to attach numerical values to said connections, therefrom to form a utilizable model of said system.

14. The knowledge engineering tool as in claim 13 wherein said component of said system is selected from the group consisting of physical and logical elements of said system.

15. The knowledge engineering tool as in claim 13 having a plurality of said interconnection cells organized according to an a priori structural knowledge of said system.

16. The knowledge engineering tool as in claim 15 wherein said a priori structural knowledge of said system is derived from information selected from the group consisting of warehouse data and expert experience.

17. The knowledge engineering tool as in claim 15 wherein said a priori structural knowledge of said system is derived from knowledge selected from the group consisting of functional, configurational, logical and heuristic structure of said system.

18. The knowledge engineering tool as in claim 15 wherein said a priori structural knowledge is derived from a member selected from the group consisting of process flow diagrams, process maps and layout drawings of said system.

19. The knowledge engineering tool as in claim 13 wherein said at least one interconnection cell has at least one output which represents an outcome of an object symbolized by said at least one interconnection cell.

20. The knowledge engineering tool as in claim 19 wherein said at least one interconnection cell has at least one input which represents an influential factor on said at least one output of said at least one interconnection cell.

21. The knowledge engineering tool as in claim 19 wherein said at least one output is selected from the group consisting of measurable output and controlled output.

22. The knowledge engineering tool as in claim 20 wherein said at least one input is selected from the group consisting of measurable input and controlled input.

23. The knowledge engineering tool as in claim 20 including at least two interconnection cells and in which said output of a first of said at least two interconnection cells is an input to a second of said at least two interconnection cells.

24. The knowledge engineering tool as in claim 20 wherein a controllable output of said first interconnection cell is a measurable input to said second interconnection cell.

25. The knowledge engineering tool as in claim 20 wherein said at least one input is an inner interrelated input.

26. The knowledge engineering tool as in claim 20 wherein said at least one input is a non obvious outside influential input.

27. A computer usable medium having a computer readable program code, the program code comprising:

a graphical symbolization unit for symbolizing objects and assumed interactions thereof by constructing at least two interconnection cells to represent components of a system whose relationship pattern is to be defined and connections therebetween to indicate said assumed interactions;

and a numerical processor associated with said graphical symbolization unit for carrying out computerized data mining amongst data representative of said system using said connections as constraints on said computerized data mining, thereby to attach numerical values to said connections to form a Knowledge-Tree map to generate a knowledge base in a data storage region of a computer.

28. The computer usable medium as in claim 27 wherein said program code is a sub-routine of a program of an automatic decision-making process.

29. The computer usable medium as in claim 28 wherein said automatic decision-making process is a part of a process control.

30. The computer usable medium as in claim 28 wherein said automatic decision-making process is suitable for a diagnostic expert system.

31. The computer usable medium as in claim 28 wherein said automatic decision-making process is suitable to trouble-shoot a process output.

32. The computer usable medium as in claim 28 wherein said automatic decision-making process is part of a microelectronics device fabrication process.

33. An automatic decision-making system comprising:

a) a data mining tool for carrying out computerized data mining to analyze available data with respect to a system to find correlations between an outcome within said system and influential factors on the outcome;

b) a Knowledge-Tree map, associated with said data mining tool for modeling a system as a series of interconnection cells with connections therebetween to represent said correlations, said Knowledge-Tree map being configured to constrain said data mining tool to correlations mapped thereon, thereby to reduce a dimension of said data mining;

c) an empirical modeler, associated with said data mining tool and said Knowledge-Tree map to use quantitative values attached to said mapped connections by said data mining tool to predict a score of said outcome; and

,

d) a decision making tool, associated with said empirical modeler, to make decisions regarding said system in accordance with said score.

34. A system as in claim 33 wherein said data mining uses statistical techniques selected from the group consisting of linear regression, nearest neighbor, clustering, process output empirical modeling (POEM), classification and regression tree (CART), chi-square automatic interaction detector (CHAID), decision trees and neural network empirical modeling.

35. A system as in claim 33 wherein said Knowledge-Tree map is a knowledge engineering tool for describing relationship pattern between plurality of objects, comprising a graphical symbolization of the objects and their relations, said graphical symbolization includes at least one interconnection cell which represents a component of a system whose said relationship pattern being described by said knowledge engineering tool.

36. A system as in claim 33 wherein said empirical modeler uses statistical techniques selected from the group consisting of linear regression, nearest neighbor, clustering, process output empirical modeling (POEM), classification and regression tree (CART), chi-square automatic interaction detector (CHAID), decision trees and neural network empirical modeling.