Systems and methods for water loss mitigation messaging

• US 10,825,096 B1
• Filed: 11/30/2016
• Issued: 11/03/2020
• Est. Priority Date: 05/23/2013
• Status: Active Grant

• Current Assignee: United Services Automobile Association
• Original Assignee: United Services Automobile Association
• Inventors: Campbell, Joseph
• Primary Examiner(s): Jacob, William J
• Application Number:
  US15/365,703
• Time in Patent Office: 1,434 Days
• Field of Search: None
• CPC Class Codes: G06N 3/08 Learning methods
• CPC Class Codes: G06N 5/01 Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks
• CPC Class Codes: G06Q 10/067 Enterprise or organisation modelling
• CPC Class Codes: G06Q 40/08 Insurance

Abstract

A computer-implemented method, includes identifying, a set of insurance policyholders that have experienced water loss and a second set of insurance policyholders that have not experienced water loss. The method also includes determining an attribute indicative of increased likelihood of future water loss using a predictive model using a percentage of the first set of insurance policyholders defining a first sample size of the first set of insurance policyholders that is smaller relative to a percentage of the second set of insurance policyholders defining the second sample size of the second set of insurance policyholders. Further, the method includes identifying at least one targeted insurance policyholder having an increased likelihood of water loss, based upon the attribute and providing a water loss mitigation strategy to the at least one targeted insurance policyholder.

Claims

1. A computer-implemented method, comprising:

identifying, from an overall set of insurance policyholders, a first set of insurance policyholders that have experienced water loss and a second set of insurance policyholders that have not experienced water loss by accessing and analyzing insurance policy information stored in a database to distinguish the first set of insurance policyholders as those that have prior water loss claims, pending water loss claims, or both and distinguish the second set of insurance policyholders as those that do not have prior water loss claims or pending water loss claims;

constructing a predictive water loss model that estimates a likelihood of a future water loss for the overall set of insurance policyholders, by;

analyzing statistics describing one or more metrics of interest using at least the insurance policy information;

data mining at least the insurance policy information to determine how variables of interest interact with one another to identify relationships between the variables of interest; and

implementing the predictive water loss model based upon the metrics of interest, the variables of interest, or both;

determining a size of a first sample of the first set of insurance policyholders and a size of a second sample of the second set of insurance policyholders such that the first sample of the first set of insurance policyholders and the second sample of the second set of insurance policyholders are balanced in a balanced data set, despite the first set of insurance policyholders being smaller than the second set of insurance policyholders, wherein the balanced data set is implemented by defining the size of the first sample as the entirety of the first set of insurance policyholders and defining the size of the second sample based upon the size of the first set of insurance policyholders;

determining an attribute indicative of increased likelihood of future water loss via the predictive water loss model using the balanced data set;

based upon the attribute, identifying at least one targeted insurance policyholder having an increased likelihood of water loss using a logistic regression that models a log-odds of water loss as opposed to non-water-loss, wherein the logistic regression is determined according to;

2. The computer-implemented method of claim 1, comprising providing the water loss mitigation strategy via an electronic message, the electronic message comprising:

a banner message presented to the at least one targeted insurance policyholder via an insurance policyholder portal, a hyperlink presented to the at least one targeted insurance policyholder directing the at least one targeted insurance policyholder to a landing page presenting various mitigation strategies, an e-mail message, or any combination thereof.

3. The computer-implemented method of claim 1, comprising:

further defining the first sample by oversampling the first set of insurance policyholders.

4. The computer-implemented method of claim 1, comprising:

further defining the second sample by randomly sampling the second set of insurance policyholders.

5. The computer-implemented method of claim 1, comprising:

identifying the at least one targeted insurance policyholder based upon the predictive water loss model, by segmenting the overall set of insurance policyholders into;

a set of high risk policyholders; and

a set of low risk policyholders; and

identifying each of the set of high risk policyholders as the at least one targeted insurance policyholder.

6. The computer-implemented method of claim 5, comprising segmenting the overall set of insurance policyholders into the set of high risk policyholders and the set of low risk policyholders by:

defining the set of high risk policyholders as those who have a risk above a predetermined threshold value; and

defining the set of low risk policyholders as those who have a risk at or below the predetermined threshold value.

7. The computer-implemented method of claim 5, comprising segmenting the overall set of insurance policyholders into the set of high risk policyholders and the set of low risk policyholders by:

defining the set of high risk policyholders as those who have a risk above an overall average of the policyholders; and

defining the set of low risk policyholders as those who have a risk at or below the overall average of the policyholders.

8. The computer-implemented method of claim 1, comprising:

determining predicted probabilities of water loss for each insurance policyholder of the overall set of insurance policyholders, according to;

9. A tangible, non-transitory, machine-readable medium, comprising machine readable instructions to:

identify, from an overall set of insurance policyholders, a first set of insurance policyholders that have experienced water loss and a second set of insurance policyholders that have not experienced water loss by accessing and analyzing insurance policy information stored in a database to distinguish the first set of insurance policyholders as those that have prior water loss claims, pending water loss claims, or both and distinguish the second set of insurance policyholders as those that do not have prior water loss claims or pending water loss claims;

construct a predictive water loss model that estimates a likelihood of a future water loss for the overall set of insurance policyholders, by;

analyzing statistics describing one or more metrics of interest using at least the insurance policy information;

data mining at least the insurance policy information to determine how variables of interest interact with one another to identify relationships between the variables of interest; and

implementing the predictive water loss model based upon the metrics of interest, the variables of interest, or both;

determine a size of a first sample of the first set of insurance policyholders and a size of a second sample of the second set of insurance policyholders such that the first sample of the first set of insurance policyholders and the second sample of the second set of insurance policyholders are balanced in a balanced data set, despite the first set of insurance policyholders being smaller than the second set of insurance policyholders, wherein the balanced data set is implemented by defining the size of the first sample as the entirety of the first set of insurance policyholders and defining the size of the second sample based upon the size of the first set of insurance policyholders;

determine an attribute indicative of increased likelihood of future water loss via the predictive water loss model using the balanced data set;

determine predicted probabilities of water loss for each insurance policyholder of the overall set of insurance policyholders, according to;

10. The machine-readable medium of claim 9, comprising instructions to:

provide the water loss mitigation strategy via an electronic message, the electronic message comprising;

a banner message presented to the at least one targeted insurance policyholder via an insurance policyholder portal, a hyperlink presented to the at least one targeted insurance policyholder directing the at least one targeted insurance policyholder to a landing page presenting various mitigation strategies, an e-mail message, or any combination thereof.

14. The machine-readable medium of claim 9, comprising instructions to:

identify the at least one targeted insurance policyholder having an increased likelihood of water loss using a logistic regression that models a log-odds of water loss as opposed to non-water-loss;

wherein the logistic regression is determined according to;

11. A computer processor configured to:

identify, from an overall set of insurance policyholders, a first set of insurance policyholders that have experienced water loss and a second set of insurance policyholders that have not experienced water loss by accessing and analyzing insurance policy information stored in a database to distinguish the first set of insurance policyholders as those that have prior water loss claims, pending water loss claims, or both and distinguish the second set of insurance policyholders as those that do not have prior water loss claims or pending water loss claims;

construct a predictive water loss model that estimates a likelihood of a future water loss for the overall set of insurance policyholders, by;

analyzing statistics describing one or more metrics of interest using at least the insurance policy information;

data mining at least the insurance policy information to determine how variables of interest interact with one another to identify relationships between the variables of interest; and

implementing the predictive water loss model based upon the metrics of interest, the variables of interest, or both;

determine a size of a first sample of the first set of insurance policyholders and a size of a second sample of the second set of insurance policyholders such that the first sample of the first set of insurance policyholders and the second sample of the second set of insurance policyholders are balanced in a balanced data set, despite the first set of insurance policyholders being smaller than the second set of insurance policyholders, by;

defining the first sample by oversampling the first set of insurance policyholders; and

defining the second sample by randomly sampling the second set of insurance policyholders;

determine an attribute indicative of increased likelihood of future water loss via the predictive water loss model using the balanced data set;

determine predicted probabilities of water loss for each insurance policyholder of the overall set of insurance policyholders, according to;

12. The computer processor of claim 11, configured to:

identify the at least one targeted insurance policyholder having an increased likelihood of water loss using a logistic regression that models a log-odds of water loss as opposed to non-water-loss;

wherein the logistic regression is determined according to;

13. The computer processor claim 11, configured to:

identify the at least one targeted insurance policyholder based upon the predictive water loss model, by segmenting the overall set of insurance policyholders into;

a set of high risk policyholders; and

a set of low risk policyholders; and

identify each of the set of high risk policyholders as the at least one targeted insurance policyholder.