STPFormer: A State-of-the-Art Pattern-Aware Spatio-Temporal Transformer for Traffic Forecasting

TL;DR

STPFormer is a novel transformer model that effectively captures complex spatio-temporal patterns for traffic forecasting, achieving state-of-the-art results through integrated modules for encoding, alignment, and fusion.

Contribution

It introduces a unified, interpretable framework with four modules that enhance spatio-temporal modeling in traffic prediction tasks.

Findings

Achieves new state-of-the-art performance on five datasets.

Modules improve temporal encoding and space-time fusion.

Model demonstrates strong generalizability and interpretability.

Abstract

Spatio-temporal traffic forecasting is challenging due to complex temporal patterns, dynamic spatial structures, and diverse input formats. Although Transformer-based models offer strong global modeling, they often struggle with rigid temporal encoding and weak space-time fusion. We propose STPFormer, a Spatio-Temporal Pattern-Aware Transformer that achieves state-of-the-art performance via unified and interpretable representation learning. It integrates four modules: Temporal Position Aggregator (TPA) for pattern-aware temporal encoding, Spatial Sequence Aggregator (SSA) for sequential spatial learning, Spatial-Temporal Graph Matching (STGM) for cross-domain alignment, and an Attention Mixer for multi-scale fusion. Experiments on five real-world datasets show that STPFormer consistently sets new SOTA results, with ablation and visualizations confirming its effectiveness and generalizability.