Pruning for Generalization: A Transfer-Oriented Spatiotemporal Graph Framework

TL;DR

The paper tackles the problem of robust cross-domain multivariate time-series forecasting on graph-structured traffic networks under data scarcity. It introduces TL-GPSTGN, a transfer-oriented framework that prunes the input graph using an entropy–correlation criterion to produce a compact, informative subgraph, which is then processed by a spatiotemporal convolutional backbone with a transfer-learning pipeline from a data-rich source to a data-sparse target. The proposed graph pruning acts as an inductive bias, reducing boundary noise and improving transferability without sacrificing essential intra-region dynamics; experiments on METR-LA, PEMS-BAY, and PEMSD7 show consistent gains in target-data-limited transfer, while single-domain performance remains competitive. The work highlights the value of structure-aware context selection in graph-based forecasting and points to future directions including incorporating exogenous signals and unsupervised pretraining to further reduce labeled-target data requirements.

Abstract

Multivariate time series forecasting in graph-structured domains is critical for real-world applications, yet existing spatiotemporal models often suffer from performance degradation under data scarcity and cross-domain shifts. We address these challenges through the lens of structure-aware context selection. We propose TL-GPSTGN, a transfer-oriented spatiotemporal framework that enhances sample efficiency and out-of-distribution generalization by selectively pruning non-optimized graph context. Specifically, our method employs information-theoretic and correlation-based criteria to extract structurally informative subgraphs and features, resulting in a compact, semantically grounded representation. This optimized context is subsequently integrated into a spatiotemporal convolutional architecture to capture complex multivariate dynamics. Evaluations on large-scale traffic benchmarks demonstrate that TL-GPSTGN consistently outperforms baselines in low-data transfer scenarios. Our findings suggest that explicit context pruning serves as a powerful inductive bias for improving the robustness of graph-based forecasting models.

Figures (5)

• Figure 1: Overview of TL-GPSTGN. The graph pruning processor (GPP) comprises an Information Entropy Analyzer (IEA), graph pruning (GP), and normalization. Each spatiotemporal convolution (ST-Conv) block follows a TCL--GCL--TCL design, i.e., two temporal convolution layers (TCL) with a graph convolution layer (GCL) in between.
• Figure 2: Training and test loss of TL-GPSTGN on metr-la. Left: 15-minute horizon. Right: 30-minute horizon. Both curves exhibit rapid early descent followed by stabilization, indicating convergence and stable generalization.
• Figure 3: Transfer-learning loss on pemsd7-m. Left: 15-minute horizon. Right: 30-minute horizon. Curves decrease steadily and converge, indicating effective adaptation to the target.
• Figure 4: Partial transfer-learning predictions on pemsd7-m. Left: 15-minute horizon. Right: 30-minute horizon. TL-GPSTGN tracks ground truth more closely than STGCN, particularly around sharp transitions and extreme values.
• Figure 5: Partial transfer-learning predictions on pems-bay. Left: 15-minute horizon. Right: 30-minute horizon. TL-GPSTGN exhibits tighter alignment with ground truth than STGCN and adapts effectively to the new network.