Weather-driven multi-category infrastructure impact forecasting

1. A computer implemented method of managing resources based on weather-related damage in a service area, the method comprising:

selecting trouble regions within a geographic service area;

assigning trouble regions to different clustered regions based on similarity of weather and weather-related damage, each of the clustered regions including at least one of the trouble regions within the geographic service area and each of the trouble regions within the geographic service area being associated with one of the clustered regions, if two or more trouble regions are assigned to a clustered region of the clustered regions, then the two or more trouble regions being similar in weather and weather-related damage;

for each clustered region of the clustered regions, training a plurality of trouble forecast models, each of the plurality of trouble forecast models being for a different particular type of damage, the training for each of the plurality of trouble forecast models including:

using training data from each trouble region within the particular clustered region, the training data including a trouble count for the particular type of damage; and

applying a scaling factor to the training data associated with each trouble region for the particular type of damage, the scaling factor being greater if the trouble count of the particular type of damage in that particular trouble region is greater than the trouble count of the particular type of damage in other trouble regions within the clustered Particular region;

applying the trouble forecast model for each trouble region within the geographic service area for the particular type of damage;

determining a trouble forecast for the geographic service area for the particular type of damage based on the trouble forecast for each of the trouble regions within the geographic service area; and

determining a job forecast for the geographic service area based on the trouble forecast for the geographic service area, wherein the managing resources is based on the job forecast for the geographic service area.

2. The computer implemented method according to claim 1 , wherein the generating the clustered regions includes clustering two or more of the trouble regions within the geographic service area or in different geographic service areas.

3. The computer implemented method according to claim 1 , wherein the training the trouble forecast model includes training a two-part model that classifies the type of damage and forecasts a number of instances of each of the types of damage.

4. The computer implemented method according to claim 1 , wherein the using the training data from each trouble region within the clustered region associated with the trouble region includes processing historical weather data associated with the trouble regions within the clustered region.

5. The computer implemented method according to claim 1 , wherein the using the training data from each trouble region within the clustered region associated with the trouble region includes obtaining training input features.

6. The computer implemented method according to claim 1 , wherein the applying the trouble forecast model includes spatially interpolating weather forecast information within the particular trouble region to a centroid of the particular trouble region to obtain interpolated data for each trouble region within the geographic service area.

7. The computer implemented method according to claim 6 , wherein the applying the trouble forecast model includes computing scoring input features from the interpolated data.

8. A system to manage resources based on weather-related damage in a service area, the system comprising:

a memory device configured to store training data; and

a processor configured to:

select trouble regions within a geographic service area;

assign trouble regions to different clustered regions based on similarity of weather and weather-related damage, each of the clustered regions including at least one of the trouble regions within the geographic service area and each of the trouble regions within the geographic service area being associated with one of the clustered regions, if two or more trouble regions are assigned to a clustered region of the clustered regions, then the two or more trouble regions being similar in weather and weather-related damage;

for each clustered region of the clustered regions, train a plurality of trouble forecast models, each of the plurality of trouble forecast models being for a different particular type of damage, the training for each of the plurality of trouble forecast models including:

using the training data from each trouble region within the particular clustered region, the training data including a trouble count for the particular type of damage; and

applying a scaling factor to the training data for each trouble region for the particular type of damage, the scaling factor being greater if the trouble count of the particular type of damage, in that particular trouble region, is greater than the trouble count of the particular type of damage in other trouble regions within the Particular clustered region;

apply the trouble forecast model for each trouble region within the geographic service area for the particular type of damage;

determine a trouble forecast for the geographic service area for the particular type of damage based on the trouble forecast for each of the trouble regions within the geographic service area; and

determine a job forecast for the geographic service area based on the trouble forecast for the geographic service area, wherein managing resources is based on the job forecast for the service area.

9. The system according to claim 8 , wherein the processor generates the clustered regions by clustering two or more of the trouble regions within the geographic service area or in different service areas.

10. The system according to claim 8 , wherein the trouble forecast model is a two-part model that classifies the type of damage and forecasts a number of instances of each of the types of damage.

11. The system according to claim 8 , wherein the processor processes historical weather data associated with the trouble regions within the clustered region.

12. The system according to claim 8 , wherein the processor obtains training input features from the training data.

13. The system according to claim 8 , wherein the processor interpolates weather forecast information within the trouble region to a centroid of the trouble region to obtain interpolated data for each trouble region within the service area.

14. The system according to claim 13 , wherein the processor computes scoring input features from the interpolated data.

15. A computer program product for managing resources based on weather-related damage in a service area, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to perform a method comprising:

selecting trouble regions within a geographic service area;

assigning trouble regions to different clustered regions based on similarity of weather and weather-related damage, each of the clustered regions including at least one of the trouble regions within the geographic service area and each of the trouble regions within the geographic service area being associated with one of the clustered regions, if two or more trouble regions are assigned to a clustered region of the clustered regions, then the two or more trouble regions being similar in weather and weather-related damage;

for each clustered region of the clustered regions, training a plurality of trouble forecast models each of the plurality of trouble forecast models being for a different particular type of damage, the training for each of the plurality of trouble forecast models including:

using training data from each trouble region within the particular clustered region, the training data including a trouble count for the particular type of damage; and

applying a scaling factor to the training data associated with each trouble region for the particular type of damage, the scaling factor being greater if the trouble count of the particular type of damage in that particular trouble region is greater than the trouble count of the particular type of damage in other trouble regions within the clustered particular region;

applying the trouble forecast model for each trouble region within the geographic service area for the particular type of damage;

determining a trouble forecast for the geographic service area for the particular type of damage based on the trouble forecast for each of the trouble regions within the geographic service area; and

determining a job forecast for the geographic service area based on the trouble forecast for the geographic service area, wherein the managing resources is based on the job forecast for the geographic service area.

16. The computer program product according to claim 15 , wherein the generating the clustered regions includes clustering two or more of the trouble regions within the geographic service area or in different geographic service areas.

17. The computer program product according to claim 15 , wherein the using the training data from each trouble region within the clustered region associated with the trouble region includes processing historical weather data associated with the trouble regions within the clustered region, and obtaining training input features.

18. The computer program product according to claim 15 , wherein the applying the trouble forecast model includes spatially interpolating weather forecast information within the trouble region to a centroid of the trouble region to obtain interpolated data for each trouble region within the service area, and computing scoring input features from the interpolated data.