US 2021/0233080 A1 · UTILIZING A TIME-DEPENDENT GRAPH CONVOLUTIONAL NEURAL NETWORK FOR FRAUDULENT TRANSACTION IDENTIFICATION

Abstract

The present disclosure relates to utilizing a graph convolutional neural network to generate similarity probabilities between pairs of digital identities associated with digital transactions based on time dependencies for use in identifying fraudulent transactions. For example, the disclosed systems can generate a transaction graph that includes nodes corresponding to digital identities. The disclosed systems can utilize a time-dependent graph convolutional neural network to generate node embeddings for the nodes based on the edge connections of the transaction graph. Further, the disclosed systems can utilize the node embeddings to determine whether a digital identity is associated with a fraudulent transaction.

Claims20

Dependent claims

In a digital medium environment for executing digital transactions, a computer-implemented method for identifying fraudulent transactions comprising:

identifying a plurality of digital identities corresponding to a plurality of digital transactions;
generating a transaction graph comprising edge connections between a plurality of nodes corresponding to the plurality of digital identities;
generating a plurality of temporal random walks between the plurality of nodes of the transaction graph;
performing a step for generating node embeddings for the nodes of the transaction graph utilizing the plurality of temporal random walks;
and
determining that a digital identity corresponds to a fraudulent transaction based on the node embeddings.

3 more elements

A non-transitory computer-readable medium storing instructions thereon that, when executed by at least one processor, cause a computing device to:

identify a plurality of digital identities corresponding to a plurality of digital transactions;
generate a transaction graph comprising edge connections between a plurality of nodes corresponding to the plurality of digital identities;
generate a plurality of temporal random walks between the plurality of nodes of the transaction graph;
utilize a time-dependent graph convolutional neural network to:
generate, based on the edge connections of the transaction graph, node embeddings corresponding to the plurality of nodes utilizing the plurality of temporal random walks;
and
determine that a digital identity from the pair of digital identities corresponds to a fraudulent transaction based on the node embeddings.

4 more elements

A system comprising:

at least one memory device comprising:
a plurality of digital identities corresponding to a plurality of digital transactions;
and
a time-dependent graph convolutional neural network trained to generate node embeddings for nodes in transaction graphs;
and
at least one server device configured to cause the system to:
generate a transaction graph comprising edge connections between a plurality of nodes corresponding to the plurality of digital identities;
determine node embeddings corresponding to the plurality of nodes utilizing the time-dependent graph convolutional neural network by, for a given node:
identifying, utilizing a plurality of temporal random walks, neighbor nodes associated with the given node based on transaction timestamps corresponding to the edge connections of the transaction graph;
and
generating a node embedding for the given node based on prior neural network layer node embeddings of the neighbor nodes;
and
determine that a digital identity corresponds to a fraudulent transaction based on the node embeddings.

10 more elements

Dependencies are inferred from the claim text.

Utilizing a time-dependent graph convolutional neural network for fraudulent transaction identification

Life of this patent

Utilizing a time-dependent graph convolutional neural network for fraudulent transaction identification

Life of this patent

Abstract

Claims20