MSHyper: Multi-Scale Hypergraph Transformer for Long-Range Time Series Forecasting

Zongjiang Shang, Ling Chen, Binqing Wu, Dongliang Cui

Abstract: Demystifying the interactions between temporal patterns of different scales is fundamental to precise long-range time series forecasting. However, previous works lack the ability to model high-order interactions. To promote more comprehensive pattern interaction modeling for long-range time series forecasting, we propose a Multi-Scale Hypergraph Transformer (MSHyper) framework. Specifically, a multi-scale feature extraction (MFE) modlue is introduced to map the input sequence into multi-scale embeddings. Then, a multi-scale hypergraph is designed to provide foundations for modeling high-order pattern interactions and a hyperedge graph is built to enhance hypergraph modeling. In addition, a tri-stage message passing (TMP) mechanism is introduced to aggregate pattern information and learn the interaction strength between temporal patterns of different scales. Extensive experimental results on eight realworld datasets demonstrate that MSHyper achieves the state-of-the-art (SOTA) performance across various settings.