System and process for multivariate adaptive regression splines classification for insurance underwriting suitable for use by an automated system

• US 7,813,945 B2
• Filed: 04/30/2003
• Issued: 10/12/2010
• Est. Priority Date: 04/30/2003
• Status: Active Grant

• Current Assignee: Pacific Life Insurance Company (Pacific Mutual Holding Company)
• Original Assignee: Genworth Financial Incorporated
• Inventors: Chakraborty, Anindya, Bonissone, Piero Patrone, Yan, Weizhong, Subbu, Rajesh Venkat, Messmer, Richard Paul
• Primary Examiner(s): Kyle; Charles R
• Assistant Examiner(s): Wood; David L
• Application Number:
  US10/425,733
• Publication Number:
  US 20040220840A1
• Time in Patent Office: 2,722 Days
• Field of Search: 705 2- 4, 700/30
• US Class Current: 705/4
• CPC Class Codes: G06Q 40/02 Banking, e.g. interest calculation or account maintenance credit or loans G06Q40/03
• CPC Class Codes: G06Q 40/08 Insurance

Abstract

A method and system for automating the decision-making process used in underwriting of insurance applications is described. While this approach is demonstrated for insurance underwriting, it is broadly applicable to diverse decision-making applications in business, commercial, and manufacturing processes. A structured methodology is used based on a multi-model parallel network of multivariate adaptive regression splines (“MARS”) models to identify the relevant set of variables and their parameters, and build a framework capable of providing automated decisions. The parameters of the MARS-based decision system are estimated from a database consisting of a set of applications with reference decisions against each. Cross-validation and development/hold-out combined with re-sampling techniques are used to build a robust set of models that minimize the error between the automated system'"'"'s decision and the expert human underwriter. Furthermore, this model building methodology can be used periodically to update and maintain the family of models if required to assure currency.

Claims

1. A process for underwriting an insurance application based on a plurality of previous insurance applications and their associated underwriting decisions, the process implemented using a computer readable medium and the process performed by a computer, the process comprising:

creating, by the computer, a plurality of binary classifiers based on a structured methodology of multivariate adaptive regression splines (“

MARS”

), where the plurality of binary classifiers are arranged in a parallel network;

identifying, by the computer, a relevant set of MARS variables and parameters based on the plurality of previous insurance applications and their associated underwriting decisions;

performing, by the computer, at least one cross-validation technique and at least one re-sampling technique on the plurality of previous insurance applications and their associated underwriting decisions;

modifying, by the computer, the plurality of binary classifiers based on the performance of the at least one cross-validation technique and the at least one re-sampling technique;

outputting, by the computer, a classification assignment for the at least one new insurance application utilizing a validated parallel network; and

performing processing on the classification assignment for the at least one new insurance application to generate an underwriting assessment of the at least one new insurance application.

2. The process according to claim 1, where the at least one cross-validation technique and the at least one re-sampling technique further comprises performing a location transformation by subtracting a normal value.

3. The process according to claim 1, where the at least one cross-validation technique and the at least one re-sampling technique further comprising normalizing the range of the plurality of applications.

4. The process according to claim 1, further comprising generating a tag for the insurance application, wherein the tag is generated based on the previous insurance applications and their associated underwriting decisions.

5. The process according to claim 1, where the at least one re-sampling technique further comprises:

partitioning data from the previous insurance applications and their associated underwriting decisions into five groups of equal size;

removing one of the five groups; and

combining the remaining four groups in a development sample.

6. The process according to claim 1, further comprising:

digitizing the insurance application and the plurality of previous insurance application underwriting decisions; and

generating a casebase of the plurality of previous insurance application underwriting decisions, where processing occurs based at least in part on the contents of the casebase.

7. The process according to claim 1, further comprising fusing the classification assignment for the at least one insurance application with at least one other classification assignment for the insurance application, where the at least one other classifier is generated by at least one other classifier.

8. A process for underwriting an insurance application based on a plurality of previous insurance applications and their associated underwriting decisions, the process implemented on a tangibly embodied computer readable medium and performed by a computer, the process comprising:

digitizing, by the computer, the insurance application and the plurality of previous insurance application underwriting decisions;

generating, by the computer, a casebase of the plurality of previous insurance application underwriting decisions;

creating, by the computer, a plurality of binary classifiers based on a structured methodology of multivariate adaptive regression splines (“

MARS”

), where the plurality of binary classifiers are arranged in a parallel network;

identifying, by the computer a relevant set of MARS variables and parameters based on the plurality of previous insurance applications and their associated underwriting decisions;

performing, by the computer, at least one cross-validation technique and at least one re-sampling technique on the plurality of previous insurance applications and their associated underwriting decisions, where the at least one re-sampling technique further comprises;

partitioning data from the previous insurance applications and their associated underwriting decisions into five groups of equal size;

removing one of the five groups; and

combining the remaining four groups in a development sample;

modifying, by the computer, the plurality of binary classifiers based on the performance of the at least one cross-validation technique and the at least one re-sampling technique;

outputting, by the computer, a classification assignment for the at least one new insurance application utilizing a validated parallel network; and

fusing, by the computer, the classification assignment for the at least one insurance application with at least one other classification assignment for the insurance application, where the at least one other classifier is generated by at least one other classifier.

9. The process according to claim 8, where the at least one cross-validation technique and the at least one re-sampling technique farther comprises performing a location transformation by subtracting a normal value.

10. The process according to claim 8, where the at least one cross-validation technique and the at least one re-sampling technique further comprising normalizing the range of the plurality of applications.

11. The process according to claim 8, further comprising generating a tag for the insurance application, wherein the tag is generated based on the previous insurance applications and their associated underwriting decisions.

12. A computer readable medium having code for causing a processor to underwrite an insurance application based on a plurality of previous insurance application underwriting decisions, the medium comprising:

code for creating a plurality of binary classifiers based on a structured methodology of multivariate adaptive regression splines (“

MARS”

), where the plurality of binary classifiers are arranged in a parallel network;

code for identifying a relevant set of MARS variables and parameters based on the plurality of previous insurance applications and their associated underwriting decisions;

code for performing at least one cross-validation technique and at least one re-sampling technique on the plurality of previous insurance applications and their associated underwriting decisions;

code for modifying the plurality of binary classifiers based on the performance of the at least one cross-validation technique and the at least one re-sampling technique; and

code for outputting a classification assignment for the at least one new insurance application utilizing a validated parallel network.

13. The medium according to claim 12, where the at least one cross-validation technique and the at least one re-sampling technique further comprises performing a location transformation by subtracting a normal value.

14. The medium according to claim 12, where the at least one cross-validation technique and the at least one re-sampling technique further comprising normalizing the range of the plurality of applications.

15. The medium according to claim 12, further comprising code for generating a tag for the insurance application, wherein the tag is generated based on the previous insurance applications and their associated underwriting decisions.

16. The medium according to claim 12, where the at least one re-sampling technique further comprises:

partitioning data from the previous insurance applications and their associated underwriting decisions into five groups of equal size;

removing one of the five groups; and

combining the remaining four groups in a development sample.

17. The medium according to claim 12, further comprising:

code for digitizing the insurance application and the plurality of previous insurance application underwriting decisions; and

code for generating a casebase of the plurality of previous insurance application underwriting decisions, where processing occurs based at least in part on the contents of the casebase.

18. The medium according to claim 12, further comprising code for fusing the classification assignment for the at least one insurance application with at least one other classification assignment for the insurance application, where the at least one other classifier is generated by at least one other classifier.

19. A computer readable medium having executable code stored on it, which, when executed by the computer, causes a processor to underwrite an insurance application based on a plurality of previous insurance application underwriting decisions, the medium comprising:

code for digitizing the insurance application and the plurality of previous insurance application underwriting decisions;

code for generating a casebase of the plurality of previous insurance application underwriting decisions;

code for creating a plurality of binary classifiers based on a structured methodology of multivariate adaptive regression splines (“

MARS”

), where the plurality of binary classifiers are arranged in a parallel network;

code for identifying a relevant set of MARS variables and parameters based on the plurality of previous insurance applications and their associated underwriting decisions;

code for performing at least one cross-validation technique and at least one re- sampling technique on the plurality of previous insurance applications and their associated underwriting decisions, where the at least one resampling technique further comprises;

partitioning data from the previous insurance applications and their associated underwriting decisions into five groups of equal size;

removing one of the five groups; and

combining the remaining four groups in a development sample;

code for modifying the plurality of binary classifiers based on the performance of the at least one cross-validation technique and at least one re-sampling technique;

code for outputting a classification assignment for the at least one new insurance application utilizing a validated parallel network; and

code for fusing the classification assignment for the at least one insurance application with at least one other classification assignment for the insurance application, where the at least one other classifier is generated by at least one other classifier.

20. The medium according to claim 19, where the at least one cross-validation technique and the at least one re-sampling technique further comprises performing a location transformation by subtracting a normal value.

21. The medium according to claim 19, where the at least one cross-validation technique and the at least one re-sampling technique further comprising normalizing the range of the plurality of applications.

22. The medium according to claim 19, further comprising code for generating a tag for the insurance application, wherein the tag is generated based on the previous insurance applications and their associated underwriting decisions.

23. A system to underwrite an insurance application based on a plurality of previous insurance application underwriting decision, the system comprising:

a classifier module for creating a plurality of binary classifiers based on a structured methodology of multivariate adaptive regression splines (“

MARS”

), where the plurality of binary classifiers are arranged in a parallel network;

an identification module for identifying a relevant set of MARS variables and parameters based on the plurality of previous insurance applications and their associated underwriting decisions;

a processor in the form of a tangibly embodied computer processor, the processor performing at least one cross-validation technique and at least one re-sampling technique on the plurality of previous insurance applications and their associated underwriting decisions, and where the classifier module modifies the plurality of binary classifiers based on the performance of the at least one cross-validation technique and at least one re-sampling technique to build a robust set of binary classifiers; and

an output module utilizing a validated parallel network for outputting a classification assignment for the at least one new insurance application.

24. The system according to claim 23, where the at least one cross-validation technique and the at least one re-sampling technique further comprises performing a location transformation by subtracting a normal value.

25. The system according to claim 23, where the at least one cross-validation technique and the at least one re-sampling technique further comprising normalizing the range of the plurality of applications.

26. The system according to claim 23, further comprising a tag module for generating a tag for the insurance application, wherein the tag is generated based on the previous insurance applications and their associated underwriting decisions.

27. The system according to claim 23, where the at least one re-sampling technique further comprises:

partitioning data from the previous insurance applications and their associated underwriting decisions into five groups of equal size;

removing one of the five groups; and

combining the remaining four groups in a development sample.

28. The system according to claim 23, further comprising:

a digitizer for digitizing the insurance application and the plurality of previous insurance application underwriting decisions; and

a storage module for creating a casebase of the plurality of previous insurance application underwriting decisions, where processing occurs based at least in part on the contents of the casebase.

29. The system according to claim 23, further comprising a fuser for fusing the classification assignment for the at least one insurance application with at least one other classification assignment for the insurance application, where the at least one other classifier is generated by at least one other classifier.

30. A system to underwrite an insurance application based on a plurality of previous insurance application underwriting decisions, the system comprising:

a digitizer for digitizing the insurance application and the plurality of previous insurance application underwriting decisions;

a storage module for creating a casebase of the plurality of previous insurance application underwriting decisions, where processing occurs based at least in part on the contents of the casebase;

a classifier module for creating a plurality of binary classifiers based on a structured methodology of multivariate adaptive regression splines (“

MARS”

), where the plurality of binary classifiers are arranged in a parallel network;

an identification module for identifying a relevant set of MARS variables and parameters based on the plurality of previous insurance applications and their associated underwriting decisions;

a processor in the form of a tangibly embodied computer processor, the processor performing at least one cross-validation technique and at least one re-sampling technique on the plurality of previous insurance applications and their associated underwriting decisions, where the classifier module modifies the plurality of binary classifiers based on the performance of the at least one cross-validation technique and at least one re-sampling technique to build a robust set of binary classifiers and where the at least one re-sampling technique further comprises;

partitioning data from the previous insurance applications and their associated underwriting decisions into five groups of equal size;

removing one of the five groups; and

combining the remaining four groups in a development sample;

an output module utilizing a validated parallel network for outputting a classification assignment for the at least one new insurance application; and

a fuser for fusing the classification assignment for the at least one insurance application with at least one other classification assignment for the insurance application, where the at least one other classifier is generated by at least one other classifier.

31. The system according to claim 30, where the at least one cross-validation technique and the at least one re-sampling technique further comprises performing a location transformation by subtracting a normal value.

32. The system according to claim 30, where the at least one cross-validation technique and the at least one re-sampling technique further comprising normalizing the range of the plurality of applications.

33. The system according to claim 30, further comprising a tag module for generating a tag for the insurance application, wherein the tag is generated based on the previous insurance applications and their associated underwriting decisions.