| CPC G06Q 40/08 (2013.01) | 20 Claims |
|
1. A computer-implemented method for determining fraudulent entities, comprising:
receiving, by one or more processors, one or more resource requests associated with a target entity;
retrieving, by the one or more processors, first characteristics data associated with one or more known fraudulent entities;
generating, by the one or more processors, a first graph based on the first characteristics data, the first graph representing (i) each of the one or more known fraudulent entities and one or more first related entities of the one or more known fraudulent entities as a first node, and (ii) each of a plurality of relationships between the one or more known fraudulent entities and the one or more first related entities as a first connection;
receiving, by the one or more processors, identification data associated with the target entity;
retrieving, by the one or more processors and using the identification data, second characteristics data associated with the target entity;
generating, by the one or more processors, a second graph based on the second characteristics data, the second graph representing (i) each of the target entity and one or more second related entities of the target entity as a second node, and (ii) each of the relationships between the target entity and the one or more second related entities as a second connection;
generating, by the one or more processors, an association score for the target entity using a first deep learning graph neural network configured to determine whether the first graph and the second graph are similar, wherein the first deep learning graph neural network is trained by:
generating a pool of connected components based on the first graph and the second graph, wherein the pool of connected components represents one or more first attributes linked to the known fraudulent entity in the first graph and the target entity in the second graph;
iteratively selecting a pair of connected components from the pool of connected components in the first graph and the second graph to compute graph similarity; and
training the first deep learning graph neural network using the computed graph similarity to determine a similarity between a pair of graphs;
determining, by the one or more processors, that the target entity is an investigative target based on the association score;
causing, by the one or more processors, a presentation of the investigative target to be displayed by a graphical user interface of a device, wherein the presentation includes a dynamic representation of a third graph indicating a third connection between (i) the target entity in the second graph, identified as the investigative target, and (ii) one or more first nodes representing the one or more known fraudulent entities in the first graph; and
generating, by the one or more processors, computer-executable instructions to invalidate the one or more resource requests associated with the target entity.
|