Sthmeo coordinates multiple traffic patterns to improve forecasting accuracy
STHMoE: Hypergraph-Enhanced Heterogeneous Dependency Coordination for LLM-Based Urban Traffic Data Forecasting
Artificial Intelligence
Summary
Predicting city traffic is hard because many sensors track different kinds of information that change over time in complex ways. The authors present STHMoE, a method that looks at traffic data from different perspectives—time, frequency, simple and complex spatial relationships—and uses a special kind of AI model called a large language model to analyze these together. This approach lets the model capture evolving traffic patterns more effectively and combines the results from different experts in a smart way. Tests on real traffic data from 10 cities showed STHMoE performs well compared to other top methods.
What this means in practice
- •For urban traffic planners: Improve traffic flow predictions by integrating diverse spatial and temporal data with evolving traffic patterns.
- •For transportation data engineers: Design traffic forecasting pipelines that adaptively combine multiple data representations using hypergraph-based expert coordination.
Authors
Jiawen Chen, Qi Shao, Yongjian Chang, Mingtong Zhou, Duxin Chen, Wenwu Yu
Abstract
Spatio-temporal traffic forecasting is a fundamental big data analytics task for intelligent transportation systems, where massive urban sensor streams exhibit heterogeneous, non-stationary, and structurally dynamic patterns. Although recent deep learning and large language model (LLM)-based methods have advanced traffic forecasting, they often remain temporally centered and lack effective coordination of temporal, spectral, pairwise spatial, and higher-order structural cues under evolving traffic regimes. To address this heterogeneous dependency coordination problem, we propose STHMoE, a Spatio-Temporal Hypergraph-Enhanced Mixture of Experts framework for urban traffic data forecasting. STHMoE decouples traffic dynamics into frequency-domain, time-domain, spatio-domain, and higher-order spatial representations, which are modeled by prompt-guided heterogeneous experts built upon a partially frozen LLM backbone. The first three experts leverage domain-specific statistical prompts, while the higher-order spatio expert uses a structural placeholder prompt and obtains dependency information from an adaptive hypergraph module. To capture evolving spatial structures in traffic data,, STHMoE jointly learns first-order graph dependencies and higher-order group interactions without predefined topologies. An entropy-aware MoE router with coefficient-of-variation load balancing adaptively fuses expert outputs while improving expert utilization and routing confidence. Experiments on 10 real-world traffic benchmarks show that STHMoE achieves competitive performance against temporal, spatio-temporal graph, and LLM-based baselines.