Leveraging Spatiotemporal Graph Neural Networks for Multi-Store Sales Forecasting

TL;DR

This paper demonstrates that spatiotemporal Graph Neural Networks significantly improve multi-store retail sales forecasting by modeling inter-store relationships, outperforming traditional methods in accuracy and stability.

Contribution

The study introduces a novel STGNN framework that effectively captures inter-store dependencies without geographic data, leading to superior forecasting performance.

Findings

STGNN achieves lowest forecasting error among tested models.

Learned adjacency matrix reveals meaningful store clusters.

Relational structure enhances forecast accuracy and robustness.

Abstract

This work evaluates the effectiveness of spatiotemporal Graph Neural Networks (GNNs) for multi-store retail sales forecasting and compares their performance against ARIMA, LSTM, and XGBoost baselines. Using weekly sales data from 45 Walmart stores, we construct a relational forecasting framework that models inter-store dependencies through a learned adaptive graph. The proposed STGNN predicts log-differenced sales and reconstructs final values through a residual path, enabling stable training and improved generalisation. Experiments show that STGNN achieves the lowest overall forecasting error, outperforming all baselines in Normalised Total Absolute Error, P90 MAPE, and variance of MAPE across stores. Analysis of the learned adjacency matrix reveals meaningful functional store clusters and high-influence nodes that emerge without geographic metadata. These results demonstrate that relational structure significantly improves forecast quality in interconnected retail environments and establishes STGNNs as a robust modelling choice for multi-store demand prediction.