Process for determining a confidence factor for insurance underwriting suitable for use by an automated system

• US 8,005,693 B2
• Filed: 06/17/2002
• Issued: 08/23/2011
• Est. Priority Date: 12/31/2001
• Status: Active Grant

• Current Assignee: Pacific Life Insurance Company (Pacific Mutual Holding Company)
• Original Assignee: Genworth Financial Incorporated
• Inventors: Cheetham, William Estel, Bonissone, Piero Patrone
• Primary Examiner(s): O'Connor; Gerald J.
• Assistant Examiner(s): Porter; Rachel L
• Application Number:
  US10/171,572
• Publication Number:
  US 20030187698A1
• Time in Patent Office: 3,354 Days
• Field of Search: 705/4
• US Class Current: 705/4
• CPC Class Codes: G06Q 10/0635 Risk analysis of enterprise or organisation activities
• CPC Class Codes: G06Q 10/10 Office automation; Time management
• CPC Class Codes: G06Q 40/08 Insurance

Abstract

A process is described for evaluating the decision-making confidence of a process and system for at least a partial underwriting of insurance policies where placement of an insurance application to an underwriting category is based on its similarity to previous insurance applications. The confidence factor computed is a measure of the correctness of the decision for a given application for insurance.

Claims

1. A process that determines confidence for an insurance application underwriting decision based on a comparison of at least one previous underwritten insurance application, the process comprising:

identifying, via an evaluation module, at least one internal parameter that may affect a conditional probability of misclassification of the insurance application;

estimating, via the evaluation module, the conditional probability of misclassification for each of the at least one identified internal parameter based at least in part on the at least one previous underwritten insurance application;

translating, via the evaluation module, the conditional probability of misclassification into a soft constraint for each of the at least one identified internal parameter, wherein the soft constraint represents a range of preference value for each of the at least one identified internal parameter;

computing a confidence factor of the underwriting decision based on the soft constraints;

defining, via the evaluation module, a run-time function to evaluate a confidence threshold for the insurance application underwriting decision based on the soft constraint; and

comparing the confidence factor with the confidence threshold to determine whether the confidence factor exceeds the confidence threshold.

2. The process according to claim 1, wherein the at least one identified internal parameter comprises at least one of:

a) a number of retrieved cases;

b) a variability of the retrieved cases;

c) a number of retrieved cases thresholded by similarity value;

d) a variability of the retrieved cases thresholded by similarity value

e) number of refined cases;

f) a variability of the refined cases

g) number of refined cases thresholded by similarity value

h) a variability of the refined cases thresholded by similarity value

i) a measure of strength of mode;

j) a number of retrieved cases weighted by similarities;

k) a variability of the retrieved cases weighted by similarities;

1) a number of refined cases weighted by similarities; and

m) a variability of the refined cases weighted for similarities.

3. The process according to claim 1, wherein the step of estimating the conditional probability of misclassification for each of the at least one identified internal parameter further comprises:

accessing a case base containing a plurality of previous underwritten insurance applications, wherein the plurality of previous underwritten insurance applications have been certified correct;

selecting one of the plurality of previous underwritten insurance applications, wherein the selected previous underwritten insurance application is defined as a probe insurance application;

identifying one or more parameters associated with the probe insurance application that may affect the conditional probability of misclassification;

determining an underwriting classification of the probe insurance application based on the remaining plurality of previous underwritten insurance applications within the case base;

comparing the underwriting classification determination with an original certified decision of the probe insurance application; and

recording the comparison and the one or more identified parameters associated with the probe insurance application.

4. The process according to claim 1, wherein the step of translating the conditional probability of misclassification into a soft constraint for each of the at least one identified internal parameter further comprises a penalty for misclassification and a reward for correct classification.

5. The process according to claim 1, wherein the step of translating the conditional probability of misclassification into a soft constraint for each of the at least one identified internal parameter further comprises an aggregating function using a weighted sum of rewards and penalties with increasing penalties for misclassifications.

6. The process according to claim 1, further comprising the step generating a soft constraint evaluation vector that contains a degree to which each of the at least one identified internal parameter satisfies the soft constraint for each of the at least one identified internal parameter.

7. The process according to claim 6, further comprising the step of computing the confidence factor based on intersection of all soft constraints evaluations contained in the soft constraint evaluation vector.

8. The process according to claim 1, wherein the insurance underwriting decision is an automated decision.

9. The process according to claim 1, further comprises eliminating one or more of the at least one identified internal parameter based at least in part on the affect of the at least one internal parameter on the conditional probability of misclassification of the insurance application, wherein the affect of the at least one internal parameter on the conditional probability of misclassification of the insurance application is evaluated via a statistical correlation analysis.

10. A process that determines confidence for an application decision based on a comparison of at least one previous application decision, the process comprising:

identifying, via an evaluation module, at least one internal parameter that may affect a conditional probability of misclassification of the application;

estimating, via the evaluation module, the conditional probability of misclassification for each of the at least one identified internal parameter based at least in part on the at least one previous application decision;

translating, via the evaluation module, the conditional probability of misclassification into a soft constraint for each of the at least one identified internal parameter, wherein the soft constraint represents a range of preference value for each of the at least one identified internal parameter;

computing a confidence factor of an underwriting decision based on the soft constraints;

defining, via the evaluation module, a run-time function to evaluate a confidence threshold for the application decision based on the soft constraint; and

comparing the confidence factor with the confidence threshold to determine whether the confidence factor exceeds the confidence threshold; and

wherein the at least one identified internal parameter includes a plurality of identified internal parameters.

11. The process according to claim 10, wherein the step of estimating the conditional probability of misclassification for each of the at least one identified internal parameter further comprises:

accessing a case base containing a plurality of previous application decisions, wherein the plurality of previous application decisions have been certified correct;

selecting one of the plurality of previous application decisions, wherein the selected previous application decision is defined as a probe application;

identifying one or more parameters associated with the probe application that may affect the conditional probability of misclassification;

determining an underwriting classification of the probe application based on the remaining plurality of previous application decisions within the case base;

comparing the underwriting classification determination with an original certified decision of the probe application; and

recording the comparison and the one or more identified parameters associated with the probe application.

12. The process according to claim 10, wherein the step of translating the conditional probability of misclassification into a soft constraint for each of the at least one identified internal parameter further comprises a penalty for misclassification and a reward for correct classification.

13. The process according to claim 10, wherein the step of translating the conditional probability of misclassification into a soft constraint for each of the at least one identified parameter further comprises an aggregating functions using a weighted sum of rewards and penalties with increasing penalties for misclassifications.

14. The process according to claim 10, further comprising the step generating a soft constraint evaluation vector that contains a degree to which each of the at least one identified internal parameter satisfies the soft constraint for each of the at least one identified internal parameter.

15. The process according to claim 14, further comprising the step of computing the confidence factor based on intersection of all soft constraints evaluations contained in the soft constraint evaluation vector.

16. The process according to claim 10, wherein the application decision is an automated application decision.

17. A process that determines confidence for an automated insurance application underwriting decision based on a comparison of at least one previous underwritten insurance application, the process comprising:

identifying, via an evaluation module, at least one internal parameter that may affect a conditional probability of misclassification of the insurance application;

estimating, via the evaluation module, the conditional probability of misclassification for the at least one identified parameter based at least in part on the at least one previous underwritten insurance application, including the sub-steps of;

a) accessing a case base containing a plurality of previous underwritten insurance applications, wherein the plurality of previous underwritten insurance applications have been certified correct;

b) selecting one of the plurality of previous underwritten insurance applications, where the selected previous underwritten insurance application is defined as a probe insurance application;

c) identifying one or more parameters associated with the probe insurance application that may affect the conditional probability of misclassification;

d) determining an underwriting classification of the probe insurance application based on the remaining plurality of previous underwritten insurance application within the case base;

e) comparing the underwriting classification determination with an original certified decision of the probe insurance application; and

f) recording the comparison and the one or more parameters associated with the probe insurance application;

translating the conditional probability of misclassification into a soft constraint for each of the at least one identified internal parameter, wherein the soft constraint represents a range of preference value for each of the at least one identified internal parameter;

computing a confidence factor of the underwriting decision based on the soft constraints;

defining a run-time function to evaluate a confidence threshold for the automated insurance application underwriting decision based on the soft constraint; and

comparing the confidence factor with the confidence threshold to determine whether the confidence factor exceeds the confidence threshold.

18. A process that determines confidence for an automated insurance application underwriting decision based on a comparison of at least one previous underwritten insurance application, the process comprising:

identifying, via an evaluation module, at least one internal parameter that may affect a conditional probability of misclassification of the insurance application;

estimating, via the evaluation module, the conditional probability of misclassification for each of the at least one identified internal parameter based at least in part on the at least one previous underwritten insurance application;

translating, via the evaluation module, the conditional probability of misclassification into a soft constraint for each of the at least one identified internal parameter, wherein the soft constraint represents a range of preference value for each of the at least one identified internal parameter and the step of translating the conditional probability of misclassification into a soft constraint further comprises a penalty for misclassification and a reward for correct classification;

computing a confidence factor of the underwriting decision based on the soft constraints;

defining, via the evaluation module, a run-time function to evaluate a confidence threshold for the automated insurance application underwriting decision based on the soft constraint; and

comparing the confidence factor with the confidence threshold to determine whether the confidence factor exceeds the confidence threshold.

19. A process that determines confidence for an automated insurance application underwriting decision based on at least one previous underwritten insurance application, the process comprising:

identifying, via an evaluation module, at least one internal parameter that is determined to potentially affect a conditional probability of misclassification of the insurance application;

estimating, via the evaluation module, the conditional probability of misclassification for each of the at least one identified internal parameter, wherein the conditional probability of misclassification is based on the at least one previous underwritten insurance application;

translating, via the evaluation module, the conditional probability of misclassification into a soft constraint for each of the at least one identified internal parameter using an aggregating function, wherein the soft constraint represents a range of preference value for each of the at least one identified internal parameter and the aggregating function uses a weighted sum of rewards and penalties with increasing penalties for misclassifications;

computing a confidence factor of the underwriting decision based on the soft constraints;

performing, via the evaluation module, a run-time function to generate a confidence factor for the automated insurance application underwriting decision, wherein the run-time function is based on the soft constraint;

comparing, via the evaluation module, the confidence factor with a confidence threshold; and

outputting, via the evaluation module, the results of the comparison.

20. A process that determines confidence for determining for an automated insurance application underwriting decision based on a comparison of at least one previous underwritten insurance application, the process comprising:

identifying, via an evaluation module, at least one internal parameter that may affect a conditional probability of misclassification of the insurance application;

estimating, via the evaluation module, the conditional probability of misclassification for each of the at least one identified internal parameter based at least in part on the at least one previous underwritten insurance application;

translating, via the evaluation module, the conditional probability of misclassification into a soft constraint for each of the at least one identified internal parameter, wherein the soft constraint represents a range of preference value for each of the at least one identified internal parameter;

computing a confidence factor of the underwriting decision based on the soft constraints;

defining, via the evaluation module, a run-time function to evaluate a confidence threshold for the automated insurance application underwriting decision based on the soft constraint;

generating, via the evaluation module, a soft constraint evaluation vector that contains a degree to which each of the at least one identified internal parameters satisfies the soft constraint for each of the at least one identified internal parameter; and

comparing the confidence factor with the confidence threshold to determine whether the confidence factor exceeds the confidence threshold.

21. The process according to claim 20, further comprising the step of computing the confidence factor based on an intersection of all soft constraints evaluations contained in the soft constraint evaluation vector.

22. A non-transitory computer-readable medium storing code that causes a processor to determine confidence for an insurance application underwriting decision based on a comparison of at least one previous underwritten insurance application, the medium comprising:

code for identifying at least one internal parameter that may affect a conditional probability of misclassification of the insurance application;

code for estimating the conditional probability of misclassification for each of the at least one identified internal parameter based at least in part on the at least one previous underwritten insurance application;

code for translating the conditional probability of misclassification into a soft constraint for each of the at least one identified internal parameter, wherein the soft constraint represents a range of preference value for each of the at least one identified internal parameter;

code for computing a confidence factor of the underwriting decision based on the soft constraints; and

code for defining a run-time function to evaluate a confidence threshold for each new query to determine the confidence factor for an insurance application underwriting decision; and

code for comparing the confidence factor with the confidence threshold to determine whether the confidence factor exceeds the confidence threshold.

23. The medium according to claim 22, where the at least one identified internal parameter comprises at least one of:

a) a number of retrieved cases;

b) a variability of the retrieved cases;

c) a number of retrieved cases thresholded by similarity value;

d) a variability of the retrieved cases thresholded by similarity value;

e) a number of refined cases;

f) a variability of the refined cases;

g) a number of refined cases thresholded by similarity value;

h) a variability of the refined cases thresholded by similarity value;

i) a measure of strength of mode;

j) a number of retrieved cases weighted by similarities;

k) a variability of the retrieved cases weighted by similarities;

1) a number of refined cases weighted by similarities; and

m) a variability of the refined cases weighted for similarities.

24. The medium according to claim 22, wherein the code for estimating the conditional probability of misclassification for each of the at least one identified internal parameter further comprises:

code for accessing a case base containing a plurality of previous underwritten insurance applications, wherein the previous underwritten insurance applications have been certified correct;

code for selecting one of the plurality of previous underwritten insurance applications, where the selected previous underwritten insurance application is defined as a probe insurance application;

code for identifying one or more parameters associated with the probe insurance application that may affect the conditional probability of misclassification;

code for determining an underwriting classification of the probe insurance application based on the remaining plurality of previous underwritten insurance application within the case base;

code for comparing the underwriting classification determination with an original certified decision of the probe insurance application; and

code for recording the comparison and the one or more parameters associated with the probe insurance application.

25. The medium according to claim 22, wherein translating the conditional probability of misclassification into a soft constraint for each of the at least one identified internal parameter further comprises a penalty for misclassification and a reward for correct classification.

26. The medium according to claim 22, wherein translating the conditional probability of misclassification into a soft constraint for each of the at least one identified internal parameter further comprises an aggregating functions using a weighted sum of rewards and penalties with increasing penalties for misclassifications.

27. The medium according to claim 22, further comprising code for generating a soft constraint evaluation vector that contains a degree to which each of the at least one identified internal parameter satisfies the soft constraint for each of the at least one identified internal parameter.

28. The medium according to claim 27, further comprising code for computing the confidence factor based on an intersection of all soft constraints evaluations contained in the soft constraint evaluation vector.

29. The medium according to claim 22, wherein the insurance application underwriting decision is an automated decision.

30. A non-transitory computer-readable medium storing code that causes a processor to determine confidence for an application decision based on a comparison of at least one previous application decision, the medium comprising:

code for identifying at least one internal parameter that may affect a conditional probability of misclassification of the application;

code for estimating the conditional probability of misclassification for each of the at least one identified internal parameters based at least in part on the at least one previous application decision;

code for translating the conditional probability of misclassification into a soft constraint for each of the at least one identified internal parameter, wherein the soft constraint represents a range of preference value for each of the at least one identified internal parameter;

code for computing a confidence factor of an underwriting decision based on the soft constraints;

code for defining a run-time function to evaluate a confidence threshold for the application decision based on the soft constraint; and

code for comparing the confidence factor with the confidence threshold to determine whether the confidence factor exceeds the confidence threshold.

31. The medium according to claim 30, wherein the code for estimating the conditional probability of misclassification for each of the at least one identified internal parameter further comprises:

code for accessing a case base containing a plurality of previous application decisions, wherein the plurality of previous application decisions have been certified correct;

code for selecting one of the plurality of previous application decisions, wherein the selected application is defined as a probe application;

code for identifying one or more parameters associated with the probe application that may affect the conditional probability of misclassification;

code for determining an underwriting classification of the probe application based on the remaining plurality of previous application decisions within the case base;

code for comparing the underwriting classification determination with an original certified decision of the probe application; and

code for recording the comparison and the one or more parameters associated with the probe application.

32. The medium according to claim 30, wherein translating the conditional probability of misclassification into a soft constraint for each of the at least one identified internal parameter further comprises a penalty for misclassification and a reward for correct classification.

33. The medium according to claim 30, wherein translating the conditional probability of misclassification into a soft constraint for each of the at least one identified parameter further comprises an aggregating functions using a weighted sum of rewards and penalties with increasing penalties for misclassifications.

34. The medium according to claim 30, further comprising code for generating a soft constraint evaluation vector that contains a degree to which each of the at least one identified internal parameter satisfies the soft constraint for each of the at least one identified internal parameter.

35. The medium according to claim 30, further comprising code for computing the confidence factor based on an intersection of all soft constraints evaluations contained in the soft constraint evaluation vector.

36. The medium according to claim 30, wherein the application decision is an automated application decision.

37. A non-transitory computer-readable medium storing code that causes a processor to determine confidence for an automated insurance application underwriting decision based on a comparison of at least one previous underwritten insurance application, the medium comprising:

code for identifying at least one internal parameter that may affect a conditional probability of misclassification of the insurance application;

code for estimating the conditional probability of misclassification for each of the at least one identified internal parameter based at least in part on the at least one previous underwritten insurance application, including;

a) code for accessing a case base containing a plurality of previous underwritten insurance applications, wherein the plurality of previous underwritten insurance applications have been certified correct;

b) code for selecting one of the plurality of previous underwritten insurance applications, wherein the selected previous underwritten insurance application is defined as a probe insurance application;

c) code for identifying at least one or more parameters of the probe insurance application that may affect the conditional probability of misclassification;

d) code for determining an underwriting classification of the probe insurance application based on the remaining plurality of previous underwritten insurance applications within the case base;

e) code for comparing an underwriting classification determination with an original certified decision of the probe insurance application; and

f) code for recording the comparison and the one or more parameters for the probe insurance application;

code for translating the conditional probability of misclassification into a soft constraint for each of the at least one identified internal parameter, wherein the soft constraint represents a range of preference value for each of the at least one identified internal parameter;

code for computing a confidence factor of the underwriting decision based on the soft constraints;

code for defining a run-time function to evaluate a confidence threshold for the automated insurance application underwriting decision based on the soft constraint; and

code for comparing the confidence factor with the confidence threshold to determine whether the confidence factor exceeds the confidence threshold.

38. A non-transitory computer-readable medium storing code that causes a processor to determine confidence for an automated insurance application underwriting decision based on a comparison of at least one previous underwritten insurance application, the medium comprising:

code for identifying at least one internal parameter that may affect a conditional probability of misclassification of the insurance application;

code for estimating the conditional probability of misclassification for each of the at least one identified internal parameter based at least in part on the at least one previous underwritten insurance application;

code for translating the conditional probability of misclassification into a soft constraint for each of the at least one identified internal parameter, wherein the soft constraint represents a range of preference value for each of the at least one identified internal parameter and the step of soft translating the conditional probability of misclassification into a soft constraint further comprises a penalty for misclassification and a reward for correct classification;

code for computing a confidence factor of the underwriting decision based on the soft constraints;

code for defining a run-time function to evaluate a confidence threshold for the automated insurance application underwriting decision based on the soft constraint; and

code for comparing the confidence factor with the confidence threshold to determine whether the confidence factor exceeds the confidence threshold.

39. A non-transitory computer-readable medium storing code that causes a processor to determine confidence for an automated insurance application underwriting decision based on a comparison of at least one previous underwritten insurance application, the medium comprising:

code for identifying at least one internal parameter that may affect a conditional probability of misclassification of the insurance application;

code for estimating the conditional probability of misclassification for each of the at least one identified internal parameter based at least in part on the at least one previous underwritten insurance application;

code for translating the conditional probability of misclassification into a soft constraint for each of the at least one identified internal parameter through an aggregating functions using a weighted sum of rewards and penalties with increasing penalties for misclassifications, wherein the soft constraint represents a range of preference value for each of the at least one identified internal parameter;

code for computing a confidence factor of the underwriting decision based on the soft constraints; and

code for defining a run-time function to evaluate a confidence threshold for the automated insurance application underwriting decision based on the soft constraint; and

code for comparing the confidence factor with the confidence threshold to determine whether the confidence factor exceeds the confidence threshold.

40. A non-transitory computer-readable medium storing code that causes a processor to determine confidence for an automated insurance application underwriting decision based on a comparison of at least one previous underwritten insurance application, the medium comprising:

code for identifying at least one internal parameter that may affect a conditional probability of misclassification of the insurance application;

code for estimating the conditional probability of misclassification for the at least one identified internal parameter based at least in part on the at least one previous underwritten insurance application;

code for translating the conditional probability of misclassification into a soft constraint for each of the at least one identified internal parameter, wherein the soft constraint represents a range of preference value for each of the at least one identified internal parameter;

code for computing a confidence factor of the underwriting decision based on the soft constraints;

code for defining a run-time function to evaluate a confidence threshold for the automated insurance application underwriting decision based on the soft constraint; and

code for generating a soft constraint evaluation vector that contains a degree to which each of the at least one identified internal parameter satisfies the soft constraint for each of the at least one identified internal parameter; and

code for comparing the confidence factor with the confidence threshold to determine whether the confidence factor exceeds the confidence threshold.

41. The medium according to claim 40, further comprising code for computing the confidence factor based on an intersection of all soft constraints evaluations contained in the soft constraint evaluation vector.

42. A process for underwriting an insurance application based on an inference of a case-based reasoning process, the case-based reasoning process comprising:

probing and retrieving, via a retrieval module, one or more previous underwritten insurance applications from a case library database; and

evaluating, via the evaluation module, the insurance application with the one or more retrieved previous underwritten insurance applications based at least in part on a confidence factor of the underwriting decision, wherein the confidence factor is determined by;

identifying at least one internal parameter that may affect a conditional probability of misclassification of the insurance application;

estimating the conditional probability of misclassification for each of the at least one identified internal parameter based at least in part on the one or more retrieved previous underwritten insurance applications;

translating the conditional probability of misclassification into a soft constraint for each of the at least one identified internal parameter, wherein the soft constraint represents a range of preference value for each of the at least one identified internal parameter;

computing the confidence factor of the underwriting decision based on the soft constraints;

defining a run-time function to evaluate a confidence threshold for the insurance application based on the soft constraint; and

comparing the confidence factor with the confidence threshold to determine whether the confidence factor exceeds the confidence threshold.

43. The process according to claim 42, further comprises eliminating one or more of the at least one identified internal parameter based at least in part on the affect of the at least one internal parameter on the conditional probability of misclassification of the insurance application, wherein the affect of the at least one internal parameter on the conditional probability of misclassification of the insurance application is evaluated via a statistical correlation analysis.