Method and system for case-based reasoning utilizing a belief network

• US 5,715,374 A
• Filed: 06/29/1994
• Issued: 02/03/1998
• Est. Priority Date: 06/29/1994
• Status: Expired due to Term

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Heckerman, David E., Breese, John S., Shaw, Gregory L.
• Primary Examiner(s): Hafiz, Tariq R.
• Application Number:
  US08/267,798
• Time in Patent Office: 1,315 Days
• Field of Search: 395/51, 395/75, 395/20, 395/60, 395/77
• US Class Current: 706/46
• CPC Class Codes: G06N 5/025 Extracting rules from data

Abstract

An improved method and system for performing case-based reasoning is provided. A belief network is utilized by the preferred case-based reasoning system for assisting a user in problem resolution. After resolving a problem of a user, the preferred embodiment of the present invention updates the probabilities in the belief network so as to provide for a more accurate problem resolution upon the next invocation of the preferred embodiment. The belief network of the preferred embodiment contains six data types relating to a problem resolution scenario. The data types utilized by the belief network of the preferred embodiment include: issues, causes, resolutions, symptoms, terms, and alternates.

Claims

1. A case-based decision support system comprising:

a storage device containing a belief network that contains knowledge obtained from observed cases so that the observed cases need not be stored, wherein each observed case is a problem resolution scenario previously resolved utilizing the belief network;

a memory containing;

an authoring component for receiving information from the user, creating the belief network from the received information, and storing the created belief network into the storage device; and

a reasoning component for receiving input from the user describing the problem encountered by the user, accessing the belief network in response to the received input from the user, and generating a resolution to the problem encountered by the user utilizing the belief network; and

a processing component for executing the authoring component and the reasoning component.

2. The case-based decision support system of claim 1 wherein the belief network contains variables for storing knowledge from observed cases comprising a plurality of types, wherein one type of the variables indicates problems of the user.

3. The case-based decision support system of claim 1 wherein the belief network contains variables for storing knowledge from observed cases comprising a plurality of types, wherein one type of the variables indicates causes for problems of the user.

4. The case-based decision support system of claim 3 wherein each variable indicating cause has terms associated with the variable indicating cause, the associated terms indicating a phrase acceptable for describing the cause of the problem.

5. The case-based decision support system of claim 4 wherein each term has alternates associated with the term, the associated alternates indicating alternative identifiers of the term.

6. The case-based decision support system of claim 1 wherein the belief network contains variables for storing knowledge from observed cases comprising a plurality of types, wherein one of the types of variables indicates symptoms for each problem of the user.

7. The case-based decision support system of claim 6 wherein each symptom variable has terms associated with the symptom variable, the associated terms indicating a phrase acceptable for describing the symptom.

8. The case-based decision support system of claim 7 wherein each term has alternates associated with the term, the associated alternates indicating alternative identifiers of the term.

9. The case-based decision support system of claim 1 wherein the belief network contains variables for storing knowledge from observed cases comprising a plurality of types, wherein one of the types of variables indicates resolutions to each problem of the user.

10. The case-based decision support system of claim 9 wherein each resolution variable has terms associated with the resolution variable, the associated terms indicating a phrase acceptable for describing a resolution.

11. The case-based decision support system of claim 10 wherein each term has alternates associated with the term, the associated alternates indicating alternative identifiers of the term.

12. The case-based decision support system of claim 1 wherein the belief network is stored in a database contained in the storage device.

20. The method of claim 1, further including the steps of:

determining whether a symptom in the first list of symptoms is observed by the user; and

generating a second list of symptoms utilizing the probabilities in the belief network in response to determining whether the symptom in the first list of symptoms is observed.

13. In a computer system, the computer system having a belief network containing probabilities for likelihood of resolutions to problems encountered by users and a memory containing a reasoner component for receiving input from the user describing a problem encountered by the user, a method for performing case-based reasoning comprising the steps of:

receiving input from the user into the reasoner component describing the problem encountered by the user;

generating a resolution to the encountered problem by the reasoner component utilizing the belief network; and

adjusting the probabilities in the belief network in response to the generated resolution.

14. The method of claim 13, further including the step of generating the belief network from user input, the belief network containing variables, arcs indicating relationships between the variables and probabilities indicating strength of the relationships between variables.

15. The method of claim 14 wherein the step of generating the belief network includes the steps of:

generating the belief network containing a plurality of variables, a problem that the user can encounter corresponding to a variable; and

for each variable corresponding to each problem of the user,

generating variables indicating causes of the problem of the user;

generating variables indicating symptoms of the problem of the user; and

generating variables indicating resolutions to the problem of the user.

16. The method of claim 15 including the step of generating terms for the cause variables, the symptom variables and the resolution variables, the generated terms for each variable indicating a phrase acceptable for describing the variable.

17. The case-based decision support system of claim 16, wherein the step of generating terms includes the step of generating alternates for each term, an alternate indicating alternative identifiers acceptable for describing the term.

18. The method of claim 13, further including the step of generating a first list of causes for the problem encountered by the user in response to receiving the input.

19. The method of claim 18, further including the steps of:

determining whether a cause in the first list of causes is observed by the user; and

generating a second list of causes utilizing the probabilities in the belief network in response to determining whether the cause in the first list of causes is observed.

21. The method of claim 13, further including the step of generating a first list of symptoms for the problem encountered by the user in response to receiving the input.

22. The method of claim 13 wherein the step of generating a resolution includes the step of generating a list of resolutions to the encountered problem, wherein the list of resolutions is ranked according to the probabilities in the belief network.

23. A case-based reasoning system for resolving a problem encountered by a user, comprising:

a belief network for generating suggested resolutions to the encountered problem of the user, the belief network having probabilities that reflect knowledge from previously observed problem resolution scenarios; and

a user interface component for interacting with the user, for receiving a description of the problem encountered by the user, for invoking the belief network to generate the suggested resolutions to the encountered problem, for updating the probabilities in the belief network to provide greater accuracy in resolving problems of the user, and for outputting the suggested resolutions to the user.

24. In a computer system, the computer system having a belief network for resolving problems of a user, the belief network having probabilities and a plurality of types of variables, the belief network storing knowledge from previously observed problem resolution scenarios, one type of variable indicating problems of a user, one type of variable indicating causes for the problems of the user, one type of variable indicating symptoms for the problems of the user, one type of variable indicating resolutions to the problems of the user, one type of variable indicating terms acceptable for describing causes, symptoms and resolutions, and one type of variable indicating alternates acceptable for identifying the terms, a method for performing case-based reasoning comprising the steps of:

invoking the belief network to suggest resolutions to a problem encountered by file user;

generating the suggested resolutions by the belief network; and

updating the probabilities in the belief network to provide greater accuracy in resolving problems of the user.

25. A computer-readable medium whose contents cause a computer system to perform case-based reasoning, the computer system having a belief network containing probabilities for likelihood of resolutions to problems encountered by users and a memory containing a reasoner component for receiving input from the user describing a problem encountered by the user, by performing the steps of:

receiving input from the user into the reasoner component describing the problem encountered by the user;

generating a resolution to the encountered problem by the reasoner component utilizing the belief network; and

adjusting the probabilities in the belief network in response to the generated resolution.

26. A computer-readable medium whose contents cause a computer system to perform case-based reasoning, the computer system having a belief network for resolving problems of a user, the belief network having probabilities and a plurality of types of variables, the belief network storing knowledge from previously observed problem resolution scenarios, by performing the steps of:

invoking the belief network to suggest resolutions to a problem encountered by the user;

generating the suggested resolutions by the belief network; and

updating the probabilities in the belief network to provide greater accuracy in resolving problems of the user.

27. A method for resolving problems by a case-based reasoning system in a computer system having a belief network with probabilities indicating likelihoods of resolutions solving problems encountered by users, comprising:

receiving a plurality of observed cases, wherein each observed case is a problem resolution scenario previously resolved by the case-based reasoning system;

for each of the received plurality of observed cases,

extracting knowledge from the observed case, wherein the knowledge influences the probabilities of the belief network;

updating the probabilities of the belief network based on the extracted knowledge; and

discarding the observed case;

receiving input from a user describing a problem encountered by the user; and

accessing the belief network and examining the probabilities to generate a resolution to the problem encountered by the user in response to receiving the input.

28. A computer-readable medium whose contents cause a case-based reasoning system in a computer system to resolve problems, the computer system having a belief network with probabilities indicating likelihoods of resolutions solving problems encountered by users, by performing the steps of:

receiving a plurality of observed cases, wherein each observed case is a problem resolution scenario previously resolved by the case-based reasoning system;

for each of the received plurality of observed cases,

extracting knowledge from the observed case, wherein the knowledge influences the probabilities of the belief network;

updating the probabilities of the belief network based on the extracted knowledge; and

discarding the observed case;

receiving input from a user describing a problem encountered by the user; and

accessing the belief network and examining the probabilities to generate a resolution to the problem encountered by the user in response to receiving the input.