Cloud-Integrated Cognitive Supply Chains: A Hybrid AI–ML Framework for Real-Time Disruption Forecasting in Retail and Manufacturing Ecosystems

Abstract

Designed to address today’s dynamic business envi- ronments, hybrid artificial intelligence–machine learning frame- works for real-time disruption forecasting are presented and examined. Their architecture comprises four key components: a data layer integrated with cloud infrastructure provisioning the real-time ingestion, preparation and storage of vast volumes of internal and external signal data; a modeling layer providing hybrid and ensembling combination of available signal sources and methods; a set of applications designed for retail and manu- facturing ecosystems; and the supporting operational principles and practices of cloud-scale AI service landscapes. Real-time data from a variety of internal and external signal sources is ingested, prepared and stored to support these applications. Two specific case studies assess how dynamic global macroeconomic conditions and extremes in the local weather pattern shape immediate consumer purchasing behaviour of certain electronic products, and how a significant disruption in physical shipments impacts the performance of the internal product-replenishment forecasting engine. One case study investigates modelling and forecasting shallow patterns in the return of finished goods from customers to the supplier, and its subsequent operational implications. The cloud-scale AI enablement model proposes the use of AI capabilities from the suppliers’ cloud environments, run as software-as-a-service AI endpoints and consumed as AI client-territory cloud services via a strong service partnership with these suppliers.