Spatial-Temporal Synchronous Graph Convolutional Networks: A New Framework for Spatial-Temporal Network Data Forecasting

Song, Chao; Lin, Youfang; Guo, Shengnan; Wan, Huaiyu · 2020 · OpenAlex-citations

DOI: 10.1609/aaai.v34i01.5438

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the challenge of forecasting spatial-temporal network data, a critical task for applications like traffic management and urban planning. The authors identify two primary limitations in existing methods: the failure to capture complex, localized spatial-temporal correlations simultaneously, and the inability to model heterogeneities across different time periods. Previous approaches typically use separate components for spatial and temporal dependencies or share modules across all time steps, which obscures distinct patterns. To resolve this, the paper proposes Spatial-Temporal Synchronous Graph Convolutional Networks (STSGCN), a framework designed to synchronously capture these correlations and heterogeneities. The STSGCN model operates by constructing localized spatial-temporal graphs that connect nodes to themselves and their neighbors across adjacent time steps. This structure allows the model to directly capture three types of influence: spatial dependencies, temporal correlations, and synchronous spatial-temporal correlations. The core component, the Spatial-Temporal Synchronous Graph Convolutional Module (STSGCM), uses graph convolutional operations in the vertex domain to aggregate features from these localized graphs. To handle heterogeneity, the model employs a Spatial-Temporal Synchronous Graph Convolutional Layer (STSGCL) that deploys multiple independent STSGCMs on different sliding windows of the time series, rather than sharing weights. Additionally, the model incorporates learnable spatial and temporal embeddings to distinguish nodes across time, a mask matrix to adjust aggregation weights based on correlation strength, and cropping operations to reduce redundancy. Experiments were conducted on four real-world datasets: PEMS03, PEMS04, PEMS07, and PEMS08, which contain traffic data from varying numbers of sensors over several months. The study compares STSGCN against several baseline methods, including traditional models like ARIMA and SVR, deep learning models like LSTM and ConvLSTM, and recent graph-based approaches such as DCRNN, STGCN, ASTGCN, and Graph WaveNet. The results demonstrate that STSGCN consistently outperforms all baseline methods, achieving state-of-the-art performance in forecasting accuracy. The significance of this work lies in its novel approach to modeling spatial-temporal data by treating spatial and temporal dimensions synchronously rather than separately. By capturing localized correlations directly and addressing heterogeneity through multi-module layers, STSGCN provides a more accurate representation of how information propagates in spatial-temporal networks. This framework offers a robust solution for improving prediction accuracy in complex networked systems, advancing the field of spatial-temporal data mining.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• anticipation