FaST: Efficient and Effective Long-Horizon Forecasting for Large-Scale Spatial-Temporal Graphs via Mixture-of-Experts

2026-01-08

arXiv.org

AI Summary

This work addresses the challenges of high computational cost, excessive memory consumption, and limited prediction accuracy faced by existing methods when performing long-horizon forecasting on large-scale spatiotemporal graphs. To this end, we propose FaST, a novel framework that introduces an adaptive graph agent attention mechanism to substantially reduce computational complexity. Furthermore, FaST incorporates a parallel Mixture-of-Experts (MoE) module based on Gated Linear Units (GLUs) to enhance model scalability and efficiency. The proposed approach enables efficient forecasting on graphs with thousands of nodes over horizons as long as 672 time steps (one week). Extensive experiments on multiple real-world datasets demonstrate that FaST significantly outperforms state-of-the-art methods, achieving high prediction accuracy while markedly reducing both computational and memory overhead.

Technology Category

• Machine Learning: Mixture of Experts (MoE)
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Large language models for search

Abstract

Spatial-Temporal Graph (STG) forecasting on large-scale networks has garnered significant attention. However, existing models predominantly focus on short-horizon predictions and suffer from notorious computational costs and memory consumption when scaling to long-horizon predictions and large graphs. Targeting the above challenges, we present FaST, an effective and efficient framework based on heterogeneity-aware Mixture-of-Experts (MoEs) for long-horizon and large-scale STG forecasting, which unlocks one-week-ahead (672 steps at a 15-minute granularity) prediction with thousands of nodes. FaST is underpinned by two key innovations. First, an adaptive graph agent attention mechanism is proposed to alleviate the computational burden inherent in conventional graph convolution and self-attention modules when applied to large-scale graphs. Second, we propose a new parallel MoE module that replaces traditional feed-forward networks with Gated Linear Units (GLUs), enabling an efficient and scalable parallel structure. Extensive experiments on real-world datasets demonstrate that FaST not only delivers superior long-horizon predictive accuracy but also achieves remarkable computational efficiency compared to state-of-the-art baselines. Our source code is available at: https://github.com/yijizhao/FaST.

Problem

• long-horizon forecasting
• spatial-temporal graph
• computational efficiency
• large-scale networks
• memory consumption

Innovation

• Mixture-of-Experts
• Spatial-Temporal Graph
• Long-Horizon Forecasting
• Graph Attention
• Gated Linear Units