Predictive bounding of combinatorial optimizations that are based on data sets acquired post-prediction through high-latency, heterogenous interfaces

Abstract

Disclosed are techniques for achieving high-volumes of low-latency price plus discount transactions that are dynamically created through transaction optimization. Aspects may include combining and optimizing different types of discounts into a single discount and dynamically controlling the single discount through a single user-facing transaction. A computer model may determine one or more transactions and sets of discounts applicable to those transactions upon which the user-facing transaction and single discount are based. Associations between discounts and items may be determined for at least some different types of discounts to apply a set of discounts including at least two different discounts of different discount types to a desired transaction for one or more items. Aspects may include determining a discount transaction comprising a plurality of constituent transactions that form a single user-facing transaction.