Graph convolutional anomaly detection
There is a need for more effective and efficient anomaly detection. This need can be addressed by, for example, solutions for performing/executing graph convolutional anomaly detection. In one example, a method includes identifying related graph database input data associated with a predictive entity; generating related graph feature data for the predictive entity; generating, based on the related graph feature data and using a graph convolutional neural network model, an anomaly detection score for the predictive entity, wherein at least a portion of the graph convolutional neural network model is trained using confirmation feedback data; performing an anomaly confirmation to generate the confirmation feedback data object for the predictive entity, and integrating the confirmation feedback data object for the predictive entity into the confirmation feedback data associated with the graph convolutional anomaly detection.
1. A computer-implemented method for performing graph convolutional anomaly detection, the computer-implemented method comprising:
identifying related graph database input data associated with a predictive entity;
generating, based at least in part on the related graph database input data, related graph feature data for the predictive entity, wherein the related graph feature data comprises a feature vector for each related graph database object of one or more related graph database objects associated with the predictive entity;
generating, based at least in part on the related graph feature data and using a graph convolutional neural network model, an anomaly detection score for the predictive entity, wherein at least a portion of the graph convolutional neural network model is trained using confirmation feedback data associated with the graph convolutional anomaly detection;
determining, based at least in part on the anomaly detection score, whether to perform anomaly confirmation with respect to the predictive entity in order to generate a confirmation feedback data object for the predictive entity;
in response to determining to perform the anomaly confirmation with respect to the predictive entity,
performing the anomaly confirmation to generate the confirmation feedback data object for the predictive entity, and
integrating the confirmation feedback data object for the predictive entity into the confirmation feedback data associated with the graph convolutional anomaly detection; and
performing one or more responsive actions based at least in part on the anomaly detection score.
2. The computer-implemented method of claim 1 , wherein identifying the graph database input data comprises:
identifying a graph database, wherein the graph database comprises a plurality of graph database objects;
selecting a related subset of the plurality of graph database objects that are deemed to be related to the predictive entity; and
generating the related graph database input data based at least in part on the related subset.
3. The computer-implemented method of claim 1 , wherein generating the related graph feature data comprises:
for each related graph database object of the one or more related graph database objects,
identifying one or more feature objects associated with the related graph database object,
parsing the one or more feature objects to generate one or more parsed feature objects,
concatenating the one or more parsed feature objects into a concatenated data object, and
processing the concatenated data object using a vectorization model to generate the feature vector for the related graph database object.
4. The computer-implemented method of claim 1 , wherein generating the anomaly detection likelihood comprises:
processing each feature vector for a related graph database object of the one or more related graph database objects using a first graph convolutional neural network of the graph convolutional neural network model to generate an anomaly presence likelihood for the predictive entity and an anomaly absence likelihood for the predictive entity; and
determining the anomaly detection score based at least in part on the anomaly presence likelihood and the anomaly absence likelihood.
5. The computer-implemented method of claim 1 , wherein:
the related graph database input data is associated with a first graph database of a plurality of graph databases from which the related graph database input data is retrieved, and
the anomaly detection score is determined based at least in part on the anomaly monitoring output of each graph database of the plurality of graph databases.
6. The computer-implemented method of claim 1 , wherein integrating the confirmation feedback data object for the predictive entity into the confirmation feedback data associated with the graph convolutional anomaly detection comprises:
monitoring a confirmation feedback stream to determine one or more feedback properties of the feedback data object;
determining, based at least in part on the one or more feedback properties, a confirmation score for the anomaly detection score;
determinizing, based at least in part on the confirmation score, a ground-truth anomaly designation for the predictive entity; and
training the graph convolutional neural network model based at least in part on the ground-truth anomaly designation.
7. The computer-implemented method of claim 1 , wherein the ground-truth anomaly designation is assigned a convolutional score based at least in part on the confirmation score.
8. The computer-implemented method of claim 1 , wherein the one or more feedback properties comprise a confirmation occurrence indicator and a confirmation latency indicator.
9. The computer-implemented method of claim 1 , wherein:
the related graph database input data is retrieved from a transactional record database; and
the anomaly detection score is a fraudulent transaction detection score.
10. The computer-implemented method of claim 1 , wherein the feature structure of the related graph feature data is determined based at least in part on an input structure of the graph convolutional neural network model.
11. An apparatus for performing graph convolutional anomaly detection, the apparatus comprising at least one processor and at least one memory including program code, the at least one memory and the program code configured to, with the processor, cause the apparatus to at least:
identify related graph database input data associated with a predictive entity;
generate, based at least in part on the related graph database input data, related graph feature data for the predictive entity, wherein the related graph feature data comprises a feature vector for each related graph database object of one or more related graph database objects associated with the predictive entity;
generate, based at least in part on the related graph feature data and using a graph convolutional neural network model, an anomaly detection score for the predictive entity, wherein at least a portion of the graph convolutional neural network model is trained using confirmation feedback data associated with the graph convolutional anomaly detection;
determine, based at least in part on the anomaly detection score, whether to perform anomaly confirmation with respect to the predictive entity in order to generate a confirmation feedback data object for the predictive entity; and
in response to determining to perform the anomaly confirmation with respect to the predictive entity,
perform the anomaly confirmation to generate the confirmation feedback data object for the predictive entity, and
integrate the confirmation feedback data object for the predictive entity into the confirmation feedback data associated with the graph convolutional anomaly detection; and
perform one or more responsive actions based at least in part on the anomaly detection score.
12. The apparatus of claim 11 , wherein identifying the graph database input data comprises:
identifying a graph database, wherein the graph database comprises a plurality of graph database objects;
selecting a related subset of the plurality of graph database objects that are deemed to be related to the predictive entity; and
generating the related graph database input data based at least in part on the related subset.
13. The apparatus of claim 11 , wherein generating the related graph feature data comprises:
for each related graph database object of the one or more related graph database objects,
identifying one or more feature objects associated with the related graph database object,
parsing the one or more feature objects to generate one or more parsed feature objects,
concatenating the one or more parsed feature objects into a concatenated data object, and
processing the concatenated data object using a vectorization model to generate the feature vector for the related graph database object.
14. The apparatus of claim 11 , wherein generating the anomaly detection likelihood comprises:
processing each feature vector for a related graph database object of the one or more related graph database objects using a first graph convolutional neural network of the graph convolutional neural network model to generate an anomaly presence likelihood for the predictive entity and an anomaly absence likelihood for the predictive entity; and
determining the anomaly detection score based at least in part on the anomaly presence likelihood and the anomaly absence likelihood.
15. The apparatus of claim 11 , wherein:
the related graph database input data is associated with a first graph database of a plurality of graph databases from which the related graph database input data is retrieved, and
the anomaly detection score is determined based at least in part on the anomaly monitoring output of each graph database of the plurality of graph databases.
16. The apparatus of claim 11 , wherein integrating the confirmation feedback data object for the predictive entity into the confirmation feedback data associated with the graph convolutional anomaly detection comprises:
monitoring a confirmation feedback stream to determine one or more feedback properties of the feedback data object;
determining, based at least in part on the one or more feedback properties, a confirmation score for the anomaly detection score;
determinizing, based at least in part on the confirmation score, a ground-truth anomaly designation for the predictive entity; and
training the graph convolutional.
17. The apparatus of claim 11 , wherein the one or more feedback properties comprise a confirmation occurrence indicator and a confirmation latency indicator.
18. The apparatus of claim 11 , wherein:
the related graph database input data is retrieved from a transactional record database; and
the anomaly detection score is a fraudulent transaction detection score.
19. The apparatus of claim 11 , wherein the feature structure of the related graph feature data is determined based at least in part on an input structure of the graph convolutional neural network model.
20. A computer program product for performing graph convolutional anomaly detection, the computer program product comprising at least one non-transitory computer-readable storage medium having computer-readable program code portions stored therein, the computer-readable program code portions configured to:
identify related graph database input data associated with a predictive entity;
generate, based at least in part on the related graph database input data, related graph feature data for the predictive entity, wherein the related graph feature data comprises a feature vector for each related graph database object of one or more related graph database objects associated with the predictive entity;
generate, based at least in part on the related graph feature data and using a graph convolutional neural network model, an anomaly detection score for the predictive entity, wherein at least a portion of the graph convolutional neural network model is trained using confirmation feedback data associated with the graph convolutional anomaly detection;
determine, based at least in part on the anomaly detection score, whether to perform anomaly confirmation with respect to the predictive entity in order to generate a confirmation feedback data object for the predictive entity;
in response to determining to perform the anomaly confirmation with respect to the predictive entity,
perform the anomaly confirmation to generate the confirmation feedback data object for the predictive entity, and
integrate the confirmation feedback data object for the predictive entity into the confirmation feedback data associated with the graph convolutional anomaly detection; and
performing one or more responsive actions based at least in part on the anomaly detection score.