Expert system method and architecture

• US 4,918,620 A
• Filed: 06/16/1988
• Issued: 04/17/1990
• Est. Priority Date: 06/16/1988
• Status: Expired due to Fees

• Current Assignee: General Electric Company
• Original Assignee: General Electric Company
• Inventors: Ulug, Mehmet E.
• Primary Examiner(s): NOT, DEFINED
• Application Number:
  US07/207,594
• Time in Patent Office: 670 Days
• Field of Search: 364/513, 364/200, 364/274.4, 364/274.6, 364/274.9, 364/274.2
• US Class Current: 706/59
• CPC Class Codes: G06N 5/022 Knowledge engineering; Knowledge acquisition

Abstract

A computer software architecture and operating method for an expert system that performs rule-based reasoning as well as a quantitative analysis, based on information provided by the user during a user session, and provides an expert system recommendation embodying the results of the quantitative analysis are disclosed. In the preferred embodiment of the invention, the expert system includes the important optional feature of modifying its reasoning process upon finding the quantitative analysis results unacceptable in comparison to predetermined acceptance criteria.

Claims

1. An expert system architecture for practice on a digital computer having memory means for storing data files, comprising:

means for storing a plurality of alphanumeric strings representing rule premises and rule conclusions as well as a rule indicator of how the premises of each rule are to be logically combined;

means for reading a user selected one of the rules stored in the memory means to enable subsequent display of the selected rule to the user.

means for displaying the selected rule to the user;

means for querying the user for a response regarding the veracity of each premise of the selected rule;

means for determining from the rule indicator a specific logic to be applied for combining the premises of the selected rule;

means for applying, upon receiving user input regarding the veracity of each rule premise, the specific logic to the selected rule premises to determine whether the specific logic is satisfied;

model means for providing an analytic model of an application being represented by the expert system, said model means comprising a plurality of subroutines that are callable for performing a quantitative analysis of said represented application;

means for querying the user for information to enable performance, using said analytic model, of a quantitative analysis associated with the selected rule;

means for calling a selected model means subroutine associated with the user selected rule;

means for providing, for the selected rule, and expert system recommendation to the user embodying results of the quantitative analysis performed by said model means;

means for comparing quantitative analysis results obtained from said selected subroutine with predetermined acceptability criteria and generating a results comparison;

means, upon finding unacceptable analytic results, for querying the user for additional information to enable performance of an additional quantitative analysis;

means for calling an additional model subroutine to perform said additional quantitative analysis; and

said system recommendation embodying the results of said additional quantitative analysis.

2. The expert system architecture of claim 1 further including:

means, upon finding unacceptable analytic results, for electing a predetermined alternate rule to substitute for said selected rule; and

means for querying the user for a response regarding the veracity of each alternate rule premise not contained in the selected rule.

3. The expert system architecture of claim 2 further including means for writing directly into said memory means to modify said selected rule in a predetermined manner to form said alternate rule.

4. The expert system architecture of claim 3 additionally including means for writing, following provision to the user of said system recommendation, directly into said memory means to reform said alternate rule into said selected rule.

5. The expert system architecture of claim 2 wherein said alternate rule is stored in said memory means, the identity of said alternate rule being predetermined in accordance with the results comparison.

6. The expert system architecture of claim 2 wherein the rules stored in said memory means can be linked to provide the conclusion of a first rule as one of the premises of a second rule.

7. The expert system architecture of claim 1 wherein the rules stored in said memory means can be linked to provide the conclusion of a first rule as one of the premises of a second rule.

8. The expert system architecture of claim 1, said model means further including neural-net means for associating a set of input data with a set of output data, said neural-net means having input and output nodes and previously determined weights between said input and output nodes for providing output data associated with user provided input data.

9. An expert system architecture for practice on a digital computer having memory means for storing data files, comprising:

system file module means for storing a plurality of alphanumeric strings representing rule premises and rule conclusions as well as a rule indicator of how the premises of each rule are to be logically combined;

system program module means including;

means for reading a user selected one of the rules stored in said file module means to enable subsequent display of the selected rule to the user;

means for displaying the selected rule to the user,

means for querying the user for a response regarding the veracity of each premise of the selected rule;

means for determining, from the rule indicator read from said file module means, a specific logic to be applied for combining the premises of the selected rule; and

means for applying, upon receiving user input regarding the veracity of each rule premise, the specific logic to be selected rule premises to determine whether the specific logic is satisfied;

model means for providing an analytic model of an application being represented by the expert system, said model means comprising a plurality of subroutines that are callable for performing a quantitative analysis of said represented application;

data module means including;

means for querying the user for information to enable performance, using said analytic model, of a quantitative analysis associated with the selected rule;

means for calling a selected model means subroutine associated with the selected rule;

means for providing, for the selected rule, an expert system recommendation to the user embodying results of the quantitative analysis performed by said model means;

means for comparing quantitative analysis results obtained from said selected subroutine with predetermined acceptability criteria and generating a results comparison;

means, upon finding unacceptable analytic results, for querying the user for additional information to enable performance of an additional quantitative analysis;

means for calling an additional model subroutine to perform the additional quantitative analysis; and

said system recommendation embodying the results of said additional quantitative analysis.

10. The expert system architecture of claim 9, said data module means further including:

means, upon finding unacceptable analytic results, for electing a predetermined alternate rule to substitute for said selected rule; and

means for querying the user for a response regarding the veracity of each alternate rule premise not contained in the selected rule.

11. The expert system architecture of claim 10, said data module means including means for writing directly into said file module means to modify said selected rule in a predetermined manner to form said alternate rule.

12. The expert system architecture of claim 11, said data module means including means for writing, following provision to the user of said system recommendation, directly into said file module means to reform said alternate rule into said selected rule.

13. The expert system architecture of claim 10 wherein said alternate rule is stored in said file module means and identified to said program module means by said data module means, the identity of said alternate rule being predetermined in accordance with the results comparison.

14. The expert system architecture of claim 10 wherein said data module means includes default information to use in lieu of information the user is unable to provide upon being queried for information to enable performance of the quantitative analysis.

15. The expert system architecture of claim 10 wherein the rules stored in said file module means can be linked to provide the conclusion of a first rule as one of the premises of a second rule.

16. The expert system architecture of claim 9 wherein said data module means includes default information to use in lieu of information the user is unable to provide upon being queried for information to enable performance of the quantitative analysis.

17. The expert system architecture of claim 9 wherein the rules stored in said file module means can be linked to provide the conclusion of a first rule as one of the premises of a second rule.

18. The expert system architecture of claim 9, said model means further including neural-net means for associating a set of input data with a set of output data, said neural-net means having input and output nodes and previously determined weights between said input and output nodes for providing output data associated with user provided input data.

19. A method for operating a digital computer to practice an expert system, said digital computer having memory means for storing data files, said method comprising the steps of:

storing in the memory means a plurality of alphanumeric strings representing rule premises and rule conclusions as well as a rule indicator of how the premises of each rule are to be logically combined;

storing in the memory means an analytic model of an application being represented by the expert system, said model comprising a plurality of subroutines that are callable for performing a quantitative analysis of said represented application and providing corresponding quantitative analysis results;

reading a user selected one of the stored rules to enable subsequent display of the selected rule to the user;

displaying the selected rule to the user;

querying the user for a response regarding the veracity of each premise of the selected rule;

determining, from the rule indicator associated with the selected rule, a specific logic to be applied for combining the premises of the selected rule;

applying, upon receiving user input regarding the veracity of each rule premise, the specific logic to the premises of the selected rule to determine whether the specific logic is satisfied;

querying the user for information to enable performance, using said analytic model, of a quantitative analysis associated with the selected rule;

calling a selected one of the model subroutines associated with the selected rule;

providing, for the selected rule, an expert system recommendation to the user embodying results of the quantitative analysis;

comparing with predetermined acceptability criteria the quantitative analysis results obtained from said selected subroutine and generating a results comparison;

querying the user, upon finding unacceptable analytic results, for additional information to enable performance of an additional quantitative analysis; and

calling an additional model subroutine to perform the additional quantitative analysis, said system recommendation embodying the results of said additional quantitative analysis.

20. The method of claim 19 including the additional steps of:

electing, upon finding unacceptable analytic results, a predetermined alternate rule to substitute for said selected rule; and

querying the user for a response regarding the veracity of each alternate rule premise not contained in the selected rule.

21. The method of claim 20 including the additional step of:

forming said alternate rule by directly modifying said selected rule in a predetermined manner in accordance with the results comparison.

22. The method of claim 21 including the further step of:

reforming said alternate rule into said selected rule following provision to the user of said system recommendation.

23. The method of claim 20 wherein said alternate rule is distinct from said selected rule;

said method comprising the additional step of;

upon electing to use said alternate rule, retrieving said alternate rule from the memory means, the identity of said alternate rule being predetermined in accordance with the results comparison.

24. The method of claim 20 wherein default information is stored in said memory means to use in lieu of information the user is unable to provide upon being queried for information to enable performance of the quantitative analysis.

25. The method of claim 19 wherein default information is stored in said memory means to use in lieu of information the user is unable to provide upon being queried for information to enable performance of the quantitative analysis.

26. The method of claim 19, said model further including neural-net means for associating a set of input data with a set of output data, said neural-net means having input and output nodes and previously determined weights between said input and output nodes for providing output data associated with user provided input data.

27. A method for operating a digital computer to practice an expert system, said digital computer having memory means for storing data files, said method comprising the steps of:

storing, in system file module means, a plurality of alphanumeric strings representing rule premises and rule conclusions as well as a rule indicator of how the premises of each rule are to be logically combined;

storing, in model means, an analytic model of an application being represented by the expert system, said model comprising a plurality of subroutines that are callable for performing a quantitative analysis of said represented application and providing corresponding quantitative analysis results;

reading, with program module means, a user selected one of the rules stored in said file module means to enable subsequent display of the selected rule to the user;

displaying to the user, with said program module means, the selected rule;

querying the user, with said program module means, for a response regarding the veracity of each premise of the selected rule;

determining within said program module means, from the rule indicator read from said file module means, a specific logic to be applied for combining the premises of the selected rule;

applying within said program module means, upon receiving user responses regarding the veracity of each rule premise, the specific logic to the premises of the selected rule to determine whether the specific logic is satisfied;

querying the user, with data module means, for information to enable performance of a quantitative analysis associated with the selected rule using said analytic model;

calling, from said data module means, selected ones of the model subroutines associated with the selected rule;

providing to the user for the selected rule, through said data module means, a system recommendation embodying results of the quantitative analysis performed by said model;

comparing with predetermined acceptability criteria, within said data module means, the quantitative analysis results obtained from said selected subroutine and generating a results comparison;

querying the user, from said data module means upon finding unacceptable analytic results, for additional information to enable performance of an additional quantitative analysis; and

calling from said data module means, an additional model subroutine to perform the additional quantitative analysis, said system recommendation embodying the results of said additional quantitative analysis.

28. The computer operating method of claim 27 including the further steps of:

electing, within said data module means upon finding unacceptable analysis results, a predetermined alternate rule to substitute for said selected rule; and

querying the user, with said program module means, for a response regarding the veracity of each alternate rule premise not contained in the selected rule.

29. The computer operating method of claim 28 including the additional step of:

forming said alternate rule by said data module means writing directly into said file module means to modify said selected rule in a predetermined manner in accordance with the results comparison.

30. The computer operating method of claim 29 including the further step of:

reforming, following provision to the user of said system recommendation, said alternate rule into said selected rule by said data module means writing directly into said file module means.

31. The expert system architecture of claim 28 wherein said alternate rule is stored in said file module means and identified to said program module means by said data module means, the identity of said alternate rule being predetermined in accordance with the results comparison.

32. The computer operating method of claim 28 wherein said data module means includes default information to use in lieu of information the user is unable to provide upon being queried for information to enable performance of the quantitative analysis.

33. The computer operating method of claim 27 wherein said data module means includes default information to use in lieu of information the user is unable to provide upon being queried for information to enable performance of the quantitative analysis.

34. The method of claim 27, said model further including neural-net means for associating a set of input data with a set of output output data, said neural-net means having input and output nodes and previously determined weights between said input and output nodes for providing output data associated with user provided input data.