Techniques for integrating proxy nodes into graph-model-based investigatory-event mappings
Methods and systems disclosed herein relate generally to generating and using graph models to perform entity-specific mappings to investigatory events. More specifically, data-reliability metrics are used to selectively use proxy nodes in graph-model trajectories during generation of the mappings.
1. A computer-implemented method comprising:
detecting a set of configurations for detecting identifiers corresponding to a particular investigatory event;
determining, using the set of configurations, a set of criteria groups for the particular investigatory event;
for each criteria group of the set of criteria groups, tagging a node in a graph structure with an identifier of the particular investigatory event, the node configured with the criteria group, the graph structure connecting a set of nodes via a set of edges, each edge of the set of edges connecting two nodes of the set of nodes, the set of nodes including a plurality of end nodes, each node of the plurality of end nodes identifying an investigatory event;
for a particular criteria group of the set of criteria groups:
identifying a data type corresponding to the particular criteria group, wherein the particular criteria group specifies an assessment of data corresponding to the data type;
accessing previous entity data corresponding to a plurality of entities, the previous entity data indicating an extent to which data corresponding to the data type was previously available in association with the plurality of entities;
defining a historical availability of the data type by:
identifying multiple data records corresponding to multiple other entities; and
determining the percentage of the multiple data records that include a value of the data type;
generating, based on the previous entity data, an estimated data-availability metric associated with a particular node configured with the particular criteria group, the estimated data-availability metric indicating an estimated extent to which the data type is available for the assessment via the particular criteria group based on the historical availability of the data type;
detecting that the estimated data-availability metric is below a predefined threshold; and
as a result of detecting that the estimated data-availability metric is below the predefined threshold, defining, for the particular node, a proxy node for the particular node, wherein the proxy node is configured with another criteria group and is defined to estimate, based on another type of data of the plurality of entity data sets, a processing result of the other criteria group;
receiving an entity data set;
generating a proxy result by evaluating the other criteria group of the proxy node using another type of data of the entity data set;
determining that an entity corresponding to the entity data set is eligible for the particular investigatory event based at least in part on the proxy result; and
outputting a result that associates an identifier of the entity and an identifier of the particular investigatory event.
2. The computer-implemented method of claim 1 , wherein the proxy node for the particular node is defined based on an output generated using an artificial-intelligence algorithm that identifies one or more data types predictive of an output of the particular node.
3. The computer-implemented method of claim 1 , further comprising, for the particular criteria group:
receiving selection data that identifies a set of entities, the set of entities including:
a first subset of the set of entities that were selected for one or more other investigatory events, wherein, for each other investigatory event of the one or more other investigatory events, an identifier of the other investigatory event was also linked to the particular node; and
a second subset of the set of entities that were not selected for the one or more other investigatory events;
retrieving a plurality of attributes that corresponds to the set of entities; and
determining, using the plurality of attributes and using a machine-learning model, that values of one or more types of attributes correspond to processing results of the particular criteria group;
wherein the other criteria group of the proxy node processes at least one of the one or more types of attributes.
4. The computer-implemented method of claim 3 , wherein the determination, using the plurality of attributes and using the machine-learning model, that values of the one or more types of attributes correspond to processing results of the particular criteria group includes:
training the machine-learning model to learn a set of machine-learning parameters, wherein one or more of the set of machine-learning parameters includes, for each of the set of attribute types, a weight or coefficient, the set of attribute types including the one or more types of attributes;
detecting, for each of the one or more types of attributes, that the weight or coefficient of the attribute exceeds a predefined threshold.
5. The computer-implemented method of claim 1 , further comprising, for each criteria group of the set of criteria groups:
querying the graph structure to determine whether the graph structure includes any existing node configured with the criteria group; and
when it is determined that the graph structure does not include any existing node configured with the criteria group, updating the graph structure to include a new node configured with the criteria group, wherein the tagging includes tagging the new node with the identifier of the particular investigatory event.
6. The computer-implemented method of claim 1 , further comprising:
generating, for each criteria group of a plurality of additional criteria groups of the set of criteria groups, an additional processing result of the criteria group using another at least part of the entity data set, wherein the determination that the entity is eligible for the investigatory event is further based on the additional processing results.
7. The computer-implemented method of claim 1 , wherein the estimated data-availability metric is based on a fraction of entities for which a value for a type of entity attribute is not included in an entity data set, the particular criteria group being configured to process the type of entity attribute.
8. The computer-implemented method of claim 1 , further comprising:
querying one or more data structures using an identifier of the data type, wherein the previous entity data was received in response to the query.
9. The computer-implemented method of claim 1 , wherein defining the proxy node for the particular node further comprises:
defining a set of additional proxy nodes for the particular node; and
using the proxy node and the set of additional proxy nodes collectively as an overall proxy to generate the proxy result for the entity.
10. The computer-implemented method of claim 1 , wherein the previous entity data includes previous trajectory data, and wherein generating the estimated data-availability metric further comprises:
identifying, based on the previous trajectory data, a set of trajectories that extend to a node configured with a criteria group that depends on an evaluation based on the data type;
identifying, based on the previous trajectory data, a subset of the set of trajectories for which the trajectory extends through the node; and
determining the estimated data-availability metric based on a size of the subset and a size of the set.
11. A system comprising:
one or more data processors; and
a non-transitory computer readable storage medium containing instructions which when executed on the one or more data processors, cause the one or more data processors to perform actions including:
detecting a set of configurations for detecting identifiers corresponding to a particular investigatory event;
determining, using the set of configurations, a set of criteria groups for the investigatory event;
for each criteria group of the set of criteria groups, tagging a node in a graph structure with an identifier of the particular investigatory event, the node configured with the criteria group, the graph structure connecting a set of nodes via a set of edges, each edge of the set of edges connecting two nodes of the set of nodes, the set of nodes including a plurality of end nodes, each node of the plurality of end nodes identifying the investigatory event;
for a particular criteria group of the set of criteria groups:
identifying a data type corresponding to the particular criteria group, wherein the particular criteria group specifies an assessment of data corresponding to the data type;
accessing previous entity data corresponding to a plurality of entities, the previous entity data indicating an extent to which data corresponding to the data type was previously available in association with the plurality of entities;
defining a historical availability of the data type by:
identifying multiple data records corresponding to multiple other entities; and
determining the percentage of the multiple data records that include a value of the data type;
generating, based on the previous entity data, an estimated data-availability metric associated with a particular node configured with the particular criteria group, the estimated data-availability metric indicating an estimated extent to which the data type is available for the assessment via the particular criteria group based on the historical availability of the data type;
detecting that the estimated data-availability metric is below a predefined threshold; and
as a result of detecting that the estimated data-availability metric is below the predefined threshold, defining, for the particular node, a proxy node for the particular node, wherein the proxy node is configured with another criteria group and is defined to estimate, based on another type of data of the plurality of entity data sets, a processing result of the other criteria group;
receiving an entity data set;
generating a proxy result by evaluating the other criteria group of the proxy node using another type of data of the entity data set;
determining that an entity corresponding to the entity data set is eligible for the particular investigatory event based at least in part on the proxy result; and
outputting a result that associates an identifier of the entity and an identifier of the particular investigatory event.
12. The system of claim 11 , wherein the proxy node for the particular node is defined based on an output generated using an artificial-intelligence algorithm that identifies one or more data types predictive of an output of the particular node.
13. The system of claim 11 , wherein the actions further include, for the particular criteria group:
receiving selection data that identifies a set of entities, the set of entities including:
a first subset of the set of entities that were selected for one or more other investigatory events, wherein, for each other investigatory event of the one or more other investigatory events, an identifier of the other investigatory event was also linked to the particular node; and
a second subset of the set of entities that were not selected for the one or more other investigatory events;
retrieving a plurality of attributes that corresponds to the set of entities; and
determining, using the plurality of attributes and using a machine-learning model, that values of one or more types of attributes correspond to processing results of the particular criteria group;
wherein the other criteria group of the proxy node processes at least one of the one or more types of attributes.
14. The system of claim 13 , wherein the determination, using the plurality of attributes and using the machine-learning model, that values of the one or more types of attributes correspond to processing results of the particular criteria group includes:
training the machine-learning model to learn a set of machine-learning parameters, wherein one or more of the set of machine-learning parameters includes, for each of the set of attribute types, a weight or coefficient, the set of attribute types including the one or more types of attributes;
detecting, for each of the one or more types of attributes, that the weight or coefficient of the attribute exceeds a predefined threshold.
15. The system of claim 11 , wherein the actions further include, for each criteria group of the set of criteria groups:
querying the graph structure to determine whether the graph structure includes any existing node configured with the criteria group; and
when it is determined that the graph structure does not include any existing node configured with the criteria group, updating the graph structure to include a new node configured with the criteria group, wherein the tagging includes tagging the new node with the identifier of the particular investigatory event.
16. The system of claim 11 , wherein the actions further include:
generating, for each criteria group of a plurality of additional criteria groups of the set of criteria groups, an additional processing result of the criteria group using another at least part of the entity data set, wherein the determination that the entity is eligible for the particular investigatory event is further based on the additional processing results.
17. The system of claim 11 , wherein the estimated data-availability metric is based on a fraction of entities for which a value for a type of entity attribute is not included in an entity data set, the particular criteria group being configured to process the type of entity attribute.
18. A computer-program product tangibly embodied in a non-transitory machine-readable storage medium, including instructions configured to cause one or more data processors to perform actions including:
detecting a set of configurations for detecting identifiers corresponding to a particular investigatory event;
determining, using the set of configurations, a set of criteria groups for the particular investigatory event;
for each criteria group of the set of criteria groups, tagging a node in a graph structure with an identifier of the particular investigatory event, the node configured with the criteria group, the graph structure connecting a set of nodes via a set of edges, each edge of the set of edges connecting two nodes of the set of nodes, the set of nodes including a plurality of end nodes, each node of the plurality of end nodes identifying the investigatory event;
for a particular criteria group of the set of criteria groups:
identifying a data type corresponding to the particular criteria group, wherein the particular criteria group specifies an assessment of data corresponding to the data type;
accessing previous entity data corresponding to a plurality of entities, the previous entity data indicating an extent to which data corresponding to the data type was previously available in association with the plurality of entities;
defining a historical availability of the data type by:
identifying multiple data records corresponding to multiple other entities; and
determining the percentage of the multiple data records that include a value of the data type;
generating, based on the previous entity data, an estimated data-availability metric associated with a particular node configured with the particular criteria group, the estimated data-availability metric indicating an estimated extent to which the data type is available for the assessment via the particular criteria group based on the historical availability of the data type;
detecting that the estimated data-availability metric is below a predefined threshold; and
as a result of detecting that the estimated data-availability metric is below the predefined threshold, defining, for the particular node, a proxy node for the particular node, wherein the proxy node is configured with another criteria group and is defined to estimate, based on another type of data of the plurality of entity data sets, a processing result of the other criteria group;
receiving an entity data set;
generating a proxy result by evaluating the other criteria group of the proxy node using another type of data of the entity data set;
determining that an entity corresponding to the entity data set is eligible for the particular investigatory event based at least in part on the proxy result; and
outputting a result that associates an identifier of the entity and an identifier of the particular investigatory event.
19. The computer-program product of claim 18 , wherein the proxy node for the particular node is defined based on an output generated using an artificial-intelligence algorithm that identifies one or more data types predictive of an output of the particular node.
20. The computer-program product of claim 18 , wherein the actions further include, for the particular criteria group:
receiving selection data that identifies a set of entities, the set of entities including:
a first subset of the set of entities that were selected for one or more other investigatory events, wherein, for each other investigatory event of the one or more other investigatory events, an identifier of the other investigatory event was also linked to the particular node; and
a second subset of the set of entities that were not selected for the one or more other investigatory events;
retrieving a plurality of attributes that corresponds to the set of entities; and
determining, using the plurality of attributes and using a machine-learning model, that values of one or more types of attributes correspond to processing results of the particular criteria group;
wherein the other criteria group of the proxy node processes at least one of the one or more types of attributes.
21. The computer-program product of claim 20 , wherein the determination, using the plurality of attributes and using the machine-learning model, that values of the one or more types of attributes correspond to processing results of the particular criteria group includes:
training the machine-learning model to learn a set of machine-learning parameters, wherein one or more of the set of machine-learning parameters includes, for each of the set of attribute types, a weight or coefficient, the set of attribute types including the one or more types of attributes;
detecting, for each of the one or more types of attributes, that the weight or coefficient of the attribute exceeds a predefined threshold.
22. The computer-program product of claim 18 , wherein the actions further include, for each criteria group of the set of criteria groups:
querying the graph structure to determine whether the graph structure includes any existing node configured with the criteria group; and
when it is determined that the graph structure does not include any existing node configured with the criteria group, updating the graph structure to include a new node configured with the criteria group, wherein the tagging includes tagging the new node with the identifier of the investigatory event.
23. The computer-program product of claim 18 , wherein the estimated data-availability metric is based on a fraction of entities for which a value for a type of entity attribute is not included in an entity data set, the particular criteria group being configured to process the type of entity attribute.