US 12,136,072 B1
Methods and systems for dynamic routing of electronic transaction messages via machine learning
Stephen E. Dinan, Austin, TX (US); James S. Osborn, Austin, TX (US); and William J. Wied, Austin, TX (US)
Assigned to Worldpay, LLC, Symmes Township, OH (US)
Filed by Worldpay, LLC, Symmes Township, OH (US)
Filed on Nov. 14, 2022, as Appl. No. 18/055,130.
Application 18/055,130 is a continuation of application No. 16/697,818, filed on Nov. 27, 2019, granted, now 12,002,015.
Int. Cl. G06Q 30/00 (2023.01); G06Q 20/04 (2012.01); G06Q 20/10 (2012.01); G06Q 20/12 (2012.01)
CPC G06Q 20/04 (2013.01) [G06Q 20/10 (2013.01); G06Q 20/12 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method for dynamic routing of electronic transactions, comprising:
receiving, by a dynamic routing service, a plurality of request messages from one or more point-of-sale (POS) devices;
processing, by the dynamic routing service, the plurality of request messages to determine routing directives associated with the plurality of request messages, wherein contextual data is retrieved based on the routing directives;
training a machine learning model based on a set of training data to determine one or more candidate processors, wherein the set of training data includes the contextual data associated with the plurality of request messages, and wherein the determining includes identifiers for the one or more candidate processors;
determining, via the trained machine learning model, a routing split for the plurality of request messages between the one or more candidate processors based on anticipated message volume during a predicted time-period;
routing, by the dynamic routing service, a varying percentage of the plurality of request messages to the one or more candidate processors based on the determined routing split for the plurality of request messages, wherein the determined routing split is fully or partially overridden upon determining the one or more candidate processors are exhibiting poor condition;
determining, via the trained machine learning model, an alternative destination processor for each of the varying percentage of the plurality of request messages whose routing split is fully or partially overridden; and
updating the identifiers of each of the varying percentage of the plurality of request messages to the alternative destination processor identifier for re-routing each of the varying percentage of the plurality of request messages to the alternative destination processor.