Machine learning system and method to manage activity notifications within utility infrastructure

1 . A computer-implemented method for automated utility infrastructure protection comprising:

receiving, by a network interface, equipment inventory data comprising utility equipment identifiers and corresponding equipment geographic coordinates stored in a utility equipment database;

generating, using a processor, a digital equipment density matrix by calculating densities of utility equipment across a plurality of geographic locations, wherein the equipment density matrix comprises a data structure mapping geographic regions to equipment density values;

processing, by the processor, structured data records comprising historical activity notifications and incident reports by identifying activity notifications and incident reports having activity locations within a predetermined geographic distance and activity timeframes and incident dates within a predetermined time span, wherein the identifying comprises correlating the activity notifications and the incident reports based on the activity locations and the activity timeframes;

generating, by the processor, combined records for each location based on the identification;

training, using a gradient boosting machine learning algorithm, a location scoring model using the combined records and applying the trained location scoring model to generate a digital location-score heatmap data structure;

receiving, through the network interface, an activity notification including an activity location;

determining, by the processor executing the trained location scoring model, a location incident score for the activity location based on the digital location-score heatmap data structure by matching the activity location to the heatmap;

determining, by the processor based on predetermined score thresholds, an activity notification disposition based on the incident score;

transmitting, via the network interface, an electronic message including the disposition;

automatically storing, in a machine learning training database, the activity notification, the location incident score, and a result of the disposition as a new historical activity record in the training dataset; and

periodically retraining the location scoring model using the gradient boosting machine learning algorithm with the updated training dataset.

2 . The method of claim 1 , wherein the disposition is selected from the group consisting of ignore the activity notification, flag the activity notification for human review, perform a physical task at the activity location, and provide instructions to an entity identified in the activity notification.

3 . The method of claim 1 , wherein the activity notification includes an entity and wherein the disposition is providing instructions to an equipment operator of the entity.

4 . The method of claim 1 , further comprising:

receiving an equipment inventory including equipment and corresponding equipment locations, at least some of the equipment locations corresponding to at least some of the plurality of locations;

generating an equipment density matrix, the equipment density matrix including a subset of the plurality of locations and a corresponding equipment density based on the equipment locations;

determining an incident score for each location in the subset by applying the location scoring model to each location in the subset; and

generating a location-score heatmap including each location in the subset with the corresponding incident score.

5 . The method of claim 4 , wherein a feature of the historical activity notifications and incident reports is location data including coordinates, wherein the location data indicates a geographical area, and wherein the method further comprises expanding the geographical area by truncating the coordinates.

6 . The method of claim 4 , wherein a feature of the historical activity notifications is an activity timeframe and a feature of the historical incident reports is an incident date, and wherein generating the combined records includes combining activity notifications and incident reports having activity locations within a predetermined distance of each other and having activity timeframes and incident dates within a predetermined time span.

7 . A non-transitory computer-readable storage medium for storing instructions for automated utility infrastructure protection executable by a processor, the instructions comprising:

receiving, by a network interface, equipment inventory data comprising utility equipment identifiers and corresponding equipment geographic coordinates stored in a utility equipment database;

generating, using the processor, a digital equipment density matrix by calculating densities of utility equipment across a plurality of geographic locations, wherein the equipment density matrix comprises a data structure mapping geographic regions to equipment density values;

processing, by the processor, structured data records comprising historical activity notifications and incident reports by identifying activity notifications and incident reports having activity locations within a predetermined geographic distance and activity timeframes and incident dates within a predetermined time span, wherein the identifying comprises correlating the activity notifications and the incident reports based on the activity locations and the activity timeframes;

generating, by the processor, combined records for each location based on the identification;

training, using a gradient boosting machine learning algorithm, a location scoring model using the combined records and applying the trained location scoring model to generate a digital location-score heatmap data structure;

receiving, through the network interface, an activity notification including an activity location;

determining, by the processor executing the trained location scoring model, a location incident score for the activity location based on the digital location-score heatmap data structure by matching the activity location to the heatmap;

determining, by the processor based on predetermined score thresholds, an activity notification disposition based on the incident score;

transmitting, via the network interface, an electronic message including the disposition;

automatically storing, in a machine learning training database, the activity notification, the location incident score, and a result of the disposition as a new historical activity record in a training dataset; and

periodically retraining the location scoring model using the gradient boosting machine learning algorithm with the updated training dataset.

8 . The computer-readable storage medium of claim 7 , wherein the disposition is selected from the group consisting of ignore the activity notification, flag the activity notification for human review, perform a physical task at the activity location, and provide instructions to an entity identified in the activity notification.

9 . The computer-readable storage medium of claim 7 , wherein the activity notification includes an entity and wherein the disposition is providing instructions to an equipment operator of the entity.

10 . The computer-readable storage medium of claim 7 , wherein the instructions further comprise:

receiving an equipment inventory including equipment and corresponding equipment locations, at least some of the equipment locations corresponding to at least some of the plurality of locations;

generating an equipment density matrix, the equipment density matrix including a subset of the plurality of locations and a corresponding equipment density based on the equipment locations;

determining an incident score for each location in the subset by applying the location scoring model to each location in the subset; and

generating a location-score heatmap including each location in the subset with the corresponding incident score.

11 . The computer-readable storage medium of claim 10 , wherein a feature of the historical activity notifications and incident reports is location data including coordinates, wherein the location data indicates a geographical area, and wherein the method further comprises expanding the geographical area by truncating the coordinates.

12 . The computer-readable storage medium of claim 10 , wherein a feature of the historical activity notifications is an activity timeframe and a feature of the historical incident reports is an incident date, and wherein generating the combined records includes combining activity notifications and incident reports having activity locations within a predetermined distance of each other and having activity timeframes and incident dates within a predetermined time span.

13 . A device comprising a processor configured to perform a method for automated utility infrastructure protection comprising:

receiving, by a network interface communicatively coupled to the processor, equipment inventory data comprising utility equipment identifiers and corresponding equipment geographic coordinates stored in a utility equipment database;

generating, using the processor, a digital equipment density matrix by calculating densities of utility equipment across a plurality of geographic locations, wherein the equipment density matrix comprises a data structure mapping geographic regions to equipment density values;

processing, by the processor, structured data records comprising historical activity notifications and incident reports by identifying activity notifications and incident reports having activity locations within a predetermined geographic distance and activity timeframes and incident dates within a predetermined time span, wherein the identifying comprises correlating the activity notifications and the incident reports based on the activity locations and the activity timeframes;

generating, by the processor, combined records for each location based on the identification;

training, using a gradient boosting machine learning algorithm, a location scoring model using the combined records and applying the trained location scoring model to generate a digital location-score heatmap data structure;

receiving, through the network interface, an activity notification including an activity location;

determining, by the processor executing the trained location scoring model, a location incident score for the activity location based on the digital location-score heatmap data structure by matching the activity location to the heatmap;

determining, by the processor based on predetermined score thresholds, an activity notification disposition based on the incident score;

transmitting, via the network interface, an electronic message including the disposition;

automatically storing, in a machine learning training database, the activity notification, the location incident score, and a result of the disposition as a new historical activity record in the training dataset; and

periodically retraining the location scoring model using the gradient boosting machine learning algorithm with the updated training dataset.

14 . The device of claim 13 , wherein the disposition is selected from the group consisting of ignore the activity notification, flag the activity notification for human review, perform a physical task at the activity location, and provide instructions to an entity identified in the activity notification.

15 . The device of claim 13 , wherein the activity notification includes an entity and wherein the disposition is providing instructions to an equipment operator of the entity.

16 . The device of claim 13 , wherein the instructions further configure the apparatus to:

receiving an equipment inventory including equipment and corresponding equipment locations, at least some of the equipment locations corresponding to at least some of the plurality of locations;

generating an equipment density matrix, the equipment density matrix including a subset of the plurality of locations and a corresponding equipment density based on the equipment locations;

determining an incident score for each location in the subset by applying the location scoring model to each location in the subset; and

generating a location-score heatmap including each location in the subset with the corresponding incident score.

17 . The device of claim 16 , wherein a feature of the historical activity notifications and incident reports is location data including coordinates, wherein the location data indicates a geographical area, and wherein the method further comprises expanding the geographical area by truncating the coordinates.

18 . The device of claim 16 , wherein a feature of the historical activity notifications is an activity timeframe and a feature of the historical incident reports is an incident date, and wherein generating the combined records include combining activity notifications and incident reports having activity locations within a predetermined distance of each other and having activity timeframes and incident dates within a predetermined time span.

19 . The computer-implemented method of claim 1 , wherein correlating the activity notifications and the incident reports comprises:

comparing the activity locations of the activity notifications with incident locations of the incident reports to identify activity notifications and incident reports within the predetermined geographic distance; and

comparing the activity timeframes of the activity notifications with the incident dates of the incident reports to identify activity notifications and incident reports within the predetermined time span,

wherein the combined records comprise activity notifications and incident reports that are within both the predetermined geographic distance and the predetermined time span.

20 . The computer-implemented method of claim 1 , wherein

the periodically retraining comprises:

retrieving the new historical activity record from the machine learning training database;

incorporating the new historical activity record into the training dataset with the historical activity notifications and incident reports; and

retraining the location scoring model using the gradient boosting machine learning algorithm based on the incorporated training dataset to improve incident score predictions for subsequent activity notifications.