| CPC G06Q 40/03 (2023.01) [G06N 3/04 (2013.01); G06N 3/08 (2013.01); G06Q 10/10 (2013.01)] | 11 Claims |
|
1. A computer-implemented method, comprising:
receiving a request for a loan from a first customer of a plurality of customers that is associated with a first node of a plurality of nodes in a network graph, the network graph being generated based on a plurality of loans, each of the plurality of loans being associated with a respective one of the plurality of customers and a respective one of a plurality of accounts;
at each of a plurality of time intervals:
identifying a plurality of attributes describing the first node;
receiving, by a graph neural network executing on a processor, a respective message from each of the plurality of nodes connected to the first node, the respective message from each of the plurality of nodes connected to the first node comprising a respective embedding vector reflecting a respective current state of a respective one of the plurality of nodes connected to the first node at a respective one of the plurality of time intervals, each of the plurality of nodes connected to the first node being associated with a respective one of the plurality of accounts;
updating, by the graph neural network executing on the processor, a plurality of weights associated with the first node based on the respective message received from each of the plurality of nodes connected to the first node at the respective one of the plurality of time intervals;
applying, by the graph neural network executing on the processor, the plurality of weights for the first node to the plurality of attributes describing the first node to compute a score reflecting credit worthiness for the customer, an optimal credit limit, or an increased credit limit relative to an initial credit limit at the respective one of the plurality of time intervals;
storing the score reflecting the credit worthiness for the customer, the optimal credit limit, or the increased credit limit as being associated with the first node at the respective one of the plurality of time intervals; and
when the score reflecting the credit worthiness for the customer exceeds a threshold, computing, by the graph neural network executing on the processor, a credit decision to approve the request for the loan at the respective one of the plurality of time intervals and approving, by the graph neural network executing on the processor, the request for the loan.
|