A Systematic Literature Review of Spatio-Temporal Graph Neural Network Models for Time Series Forecasting and Classification

TL;DR

This systematic literature review synthesizes 366 studies on spatio-temporal GNNs applied to time series forecasting and classification across energy, environment, finance, health, mobility, and more. It classifies models into recurrent, convolutional, and attentional GNNs, with extensive discussion of graph structure design (predefined vs learned) and evaluation practices, highlighting a trend toward self-learned graphs and convolutional spatial aggregation. The authors compile comprehensive tables and an interactive GitHub resource to compare benchmarks, datasets, and results, revealing that while GNN-based approaches often outperform baselines in many domains, the field suffers from fragmented benchmarks and limited reproducibility. They advocate for standardized datasets, transparent reporting, and more interpretable, scalable models to advance cross-domain applicability and practical deployment. The study also points to future directions such as probabilistic forecasting, integration with transformers, and improved explainability. Overall, the work provides a valuable, multi-domain reference that can guide researchers in model selection, benchmarking, and methodological development in spatio-temporal GNNs.

Abstract

In recent years, spatio-temporal graph neural networks (GNNs) have attracted considerable interest in the field of time series analysis, due to their ability to capture, at once, dependencies among variables and across time points. The objective of this systematic literature review is hence to provide a comprehensive overview of the various modeling approaches and application domains of GNNs for time series classification and forecasting. A database search was conducted, and 366 papers were selected for a detailed examination of the current state-of-the-art in the field. This examination is intended to offer to the reader a comprehensive review of proposed models, links to related source code, available datasets, benchmark models, and fitting results. All this information is hoped to assist researchers in their studies. To the best of our knowledge, this is the first and broadest systematic literature review presenting a detailed comparison of results from current spatio-temporal GNN models applied to different domains. In its final part, this review discusses current limitations and challenges in the application of spatio-temporal GNNs, such as comparability, reproducibility, explainability, poor information capacity, and scalability. This paper is complemented by a GitHub repository at https://github.com/FlaGer99/SLR-Spatio-Temporal-GNN.git providing additional interactive tools to further explore the presented findings.

Figures (9)

• Figure 1: Overview of the structure of the SLR.
• Figure 2: PRISMA flowchart summarizing the identification and selection process, adapted from PRISMA_statement. The flowchart is structured vertically into three key phases: identification (collection of records from databases, and removal of duplicates and non-top-tier venues), screening (assessment for eligibility and exclusion of out-of-scope studies), and inclusion (inclusion of papers in the review). The boxes on the right indicate the number of studies excluded at each stage.
• Figure 3: Number of journal and conference papers over time across the different groups.
• Figure 4: Pie chart of the distribution of corresponding authors in different countries.
• Figure 5: Map of the complete collaboration network.

Theorems & Definitions (5)

• Definition 1: Time series
• Definition 2: Graph
• Definition 3: Spatio-temporal graph
• Definition 4: Adjacency matrix
• Definition 5: Degree matrix