METHOD AND SYSTEM FOR CREATING A DYNAMIC SYSTEMS BASED HYBRID MODEL FOR REASONING SYSTEMS

• US 20120158640A1
• Filed: 12/17/2010
• Published: 06/21/2012
• Est. Priority Date: 12/17/2010
• Status: Active Grant

• Current Assignee: General Electric Company
• Original Assignee: General Electric Company
• Inventors: Repoff, Thomas Paul, Clark, Michael Craig, Venditti, Paul
• Granted Patent: US 8,533,147 B2
• Time in Patent Office: Days
• US Class Current: 706/59
• CPC Class Codes: G06N 5/04 Inference or reasoning models

Abstract

A method for creating a dynamic systems based hybrid model for reasoning systems is described. The method receives, at a server, a plurality of expert identified deal attributes. The method presents a conjoint analysis questionnaire for input by one or more respondents. The conjoint analysis questionnaire includes questions based on the plurality of expert identified deal attributes. The method receives responses of the one or more respondents to the conjoint analysis questionnaire. The method then defines an expert reasoning model based on the responses of the one or more respondents to the conjoint analysis questionnaire. The method accesses historical deal information of one or more deals, the historical deal information including one or more deals, each deal including a plurality of expert identified deal attributes and a part-worth associated with each of the deal attributes. Finally, the method validates the expert reasoning model based on the historical deal information.

Claims

1. A method comprising:

receiving, at a server, a plurality of expert identified deal attributes;

presenting a conjoint analysis questionnaire for input by one or more respondents, wherein the conjoint analysis questionnaire includes questions based on the plurality of expert identified deal attributes;

receiving, at the server, responses of the one or more respondents to the conjoint analysis questionnaire;

defining an expert reasoning model based on the responses of the one or more respondents to the conjoint analysis questionnaire;

accessing historical deal information of one or more deals, the historical deal information including one or more deals, each deal including a plurality of expert identified deal attributes and a part-worth associated with each of the deal attributes; and

validating the expert reasoning model based on the historical deal information.

2. The method of claim 1 wherein presenting the conjoint analysis questionnaire further comprises:

presenting one or more deal attribute rating questions to prompt the respondent to assign importance levels to the plurality of expert identified deal attributes;

selecting a predetermined number of critical deal attributes from the plurality of expert identified deal attributes, based on responses to the one or more deal attribute rating questions;

presenting one or more trade-off questions based on the selected critical deal attributes; and

presenting one or more calibration questions, to prompt the respondent to provide a scaling factor.

3. The method of claim 2 wherein defining the expert reasoning model further comprises:

for each respondent, computing part-worths for each of the critical deal attributes based on the responses to the conjoint analysis questionnaire;

for each respondent, computing a utility function based on the computed part-worths and the critical deal attributes; and

creating the expert reasoning model based on the utility functions of each respondent.

4. The method of claim 3 wherein defining the expert reasoning model further comprises:

for each respondent, determining a measure of temporal stability, based on the computed part-worth for each of the critical deal attributes, and an expected part-worth for each of the critical deal attributes; and

modifying the utility function based on the measure of temporal stability.

5. The method of claim 4 wherein the measure of temporal stability is determined using one or more of an apriori sign violation method, a dynamic Bayesian linear method, a regression tree method, a validity correlation method, Rasch modeling, and an association rule mining method.

6. The method of claim 2 wherein defining the expert reasoning model further comprises scaling the part-worths for each of the critical deal attributes based on the responses to the one or more calibration questions.

7. A system comprising:

at least one processor;

computer program code stored in a computer readable storage medium, wherein the computer program code, when executed, is operative to cause the at least one processor to;

receive a plurality of expert identified deal attributes;

present a conjoint analysis questionnaire for input by one or more respondents, wherein the conjoint analysis questionnaire includes questions based on the plurality of expert identified deal attributes;

receive responses of the one or more respondents to the conjoint analysis questionnaire;

define an expert reasoning model based on the responses of the one or more respondents to the conjoint analysis questionnaire;

access historical deal information of one or more deals, the historical deal information including one or more deals, each deal including a plurality of expert identified deal attributes and a part-worth associated with each of the deal attributes; and

validate the expert reasoning model based on the historical deal information.

8. The system of claim 7 the computer program code is further operative to cause the at least one processor to:

present one or more deal attribute rating questions to prompt the respondent to assign importance levels to the plurality of expert identified deal attributes;

select a predetermined number of critical deal attributes from the plurality of expert identified deal attributes, based on responses to the one or more deal attribute rating questions;

present one or more trade-off questions based on the selected critical deal attributes; and

present one or more calibration questions, to prompt the respondent to provide a scaling factor.

9. The system of claim 8 the computer program code is further operative to cause the at least one processor to:

for each respondent, compute part-worths for each of the critical deal attributes based on the responses to the conjoint analysis questionnaire;

for each respondent, compute a utility function based on the computed part-worths and the critical deal attributes; and

create the expert reasoning model based on the utility functions of each respondent.

10. The system of claim 9 the computer program code is further operative to cause the at least one processor to:

for each respondent, determine a measure of temporal stability, based on the computed part-worth for each of the critical deal attributes, and an expected part-worth for each of the critical deal attributes; and

modify the utility function based on the measure of temporal stability.

11. The system of claim 10 wherein the measure of temporal stability is determined using one or more of an apriori sign violation method, a dynamic Bayesian linear method, a regression tree method, a validity correlation method, Rasch modeling, and an association rule mining method.

12. The system of claim 8 wherein defining the expert reasoning model further comprises scaling the part-worths for each of the critical deal attributes based on the responses to the one or more calibration questions.

13. A computer program product comprising a non-transitory computer readable medium encoded with computer-executable instructions, wherein the computer-executable instructions, when executed, cause one or more processors to:

receive a plurality of expert identified deal attributes;

present a conjoint analysis questionnaire for input by one or more respondents, wherein the conjoint analysis questionnaire includes questions based on the plurality of expert identified deal attributes;

receive responses of the one or more respondents to the conjoint analysis questionnaire;

define an expert reasoning model based on the responses of the one or more respondents to the conjoint analysis questionnaire;

access historical deal information of one or more deals, the historical deal information including one or more deals, each deal including a plurality of expert identified deal attributes and a part-worth associated with each of the deal attributes; and

validate the expert reasoning model based on the historical deal information.

14. The computer program product of claim 13 further comprising computer executable instructions to cause the one or more processors to:

present one or more deal attribute rating questions to prompt the respondent to assign importance levels to the plurality of expert identified deal attributes;

select a predetermined number of critical deal attributes from the plurality of expert identified deal attributes, based on responses to the one or more deal attribute rating questions;

present one or more trade-off questions based on the selected critical deal attributes; and

present one or more calibration questions, to prompt the respondent to provide a scaling factor.

15. The computer program product of claim 14 further comprising computer executable instructions to cause the one or more processors to:

compute part-worths for each of the critical deal attributes based on the responses to the conjoint analysis questionnaire, for each respondent;

compute a utility function based on the computed part-worths and the critical deal attributes, for each respondent; and

create the expert reasoning model based on the utility functions of each respondent.

16. The computer program product of claim 15 further comprising computer executable instructions to cause the one or more processors to:

determine a measure of temporal stability for each respondent based on the computed part-worth for each of the critical deal attributes, and an expected part-worth for each of the critical deal attributes; and

modify the utility function based on the measure of temporal stability.

17. The computer program product of claim 16 wherein the measure of temporal stability is determined using one or more of an apriori sign violation method, a dynamic Bayesian linear method, a regression tree method, a validity correlation method, Rasch modeling, and an association rule mining method.

18. The computer program product of claim 15 further comprising computer executable instructions to cause the one or more processors to scale the part-worths for each of the critical deal attributes based on the responses to the one or more calibration questions.