Temporal Graph Neural Networks for Multi-Product Time Series Forecasting

Author(s): Shenggang Li

Originally published on Towards AI.

Modeling Cross-Series Dependencies and Temporal Dynamics in Retail Supply-Chain Data

This paper uses a supply-chain scenario — forecasting daily sales for multiple products — to explore Graph Neural Networks (GNNs) and their temporal extensions from first principles. In a network of related SKUs, a spike in one often ripples into the others. We demonstrate how to learn a sparse influence graph, apply graph convolutions to blend neighbor information, and then layer in temporal convolutions to capture evolving patterns — unpacking both the mechanics and the math behind each step.

This article explores the use of Temporal Graph Neural Networks (TGNNs) for forecasting multiple product demands in a supply chain context. It details how TGNNs effectively learn the influence of related products on one another, handle temporal dynamics, and improve forecasting accuracy. With practical insights and examples, the piece provides a comprehensive guide to implementing TGNNs, demonstrating their superior performance over traditional forecasting methods by accurately capturing complex interdependencies and seasonal trends.

Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI