| CPC G06Q 20/4016 (2013.01) [G06N 3/02 (2013.01); G06N 5/02 (2013.01); G06Q 20/4014 (2013.01)] | 20 Claims |
|
1. A method of detecting and responding to fraudulent financial activity using a knowledge graph, comprising:
retrieving the knowledge graph, wherein the knowledge graph includes a set of nodes corresponding to a plurality of financial transactions;
the set of nodes in the knowledge graph having been generated from non-homogenous data corresponding to various events received from across a plurality of different communication channels;
the plurality of communication channels including data originating from a voice-based communication channel related to one or more natural language conversations between a user and an agent of a call center in combination with one or more channels-selected from the group consisting of text data, webpage data, mobile application data, form data, interactions between the user and a webpage, interactions between an agent of the user and a webpage, and combinations thereof;
generating an embedding of the knowledge graph within an embedding space;
analyzing the embedding of the knowledge graph and detecting a fraudulent pattern within the embedding of the knowledge graph;
wherein analyzing the embedding of the knowledge graph includes identifying a subset of nodes, out of the set of nodes in the knowledge graph, that are descriptive of transactions performed at a common merchant;
wherein the step of identifying a subset of nodes that are descriptive of transaction performed at the common merchant is iterative and uses information from previous embeddings, as a result of the steps of retrieving the knowledge graph, generating an embedding of the knowledge graph, and analyzing the embedding of the knowledge graph being repeated using sequential data; and
automatically taking a fraud limiting action in response to a dynamic system detecting efficiently the fraudulent pattern, the fraud limiting action including sending a message to a mobile computing device associated with the user;
wherein the fraud limiting action further includes taking a first fraud limiting action when a number of disputed transactions at the common merchant is below a first threshold, taking a second fraud limiting action when the number of disputed transactions at the common merchant is greater than the first threshold and less than a second threshold, taking a third fraud limiting action when the number of disputed transactions at the common merchant is greater than the second threshold and less than a third threshold, and taking a fourth fraud limiting action when the number of disputed transactions at the common merchant is greater than the third threshold.
|