Edge-Varying Fourier Graph Networks for Multivariate Time Series Forecasting

TL;DR

This paper introduces Edge-Varying Fourier Graph Networks (EV-FGN), a novel approach that adaptively models dynamic variable correlations in multivariate time series using frequency domain graph convolution, leading to improved forecasting accuracy.

Contribution

The paper proposes EV-FGN with an adaptive supra-graph and Fourier-based graph convolution, capturing time-varying dependencies more effectively than static graph methods.

Findings

EV-FGN outperforms state-of-the-art methods on seven real-world datasets.

The frequency domain approach improves computational efficiency.

Adaptive supra-graph captures dynamic correlations better than static graphs.

Abstract

The key problem in multivariate time series (MTS) analysis and forecasting aims to disclose the underlying couplings between variables that drive the co-movements. Considerable recent successful MTS methods are built with graph neural networks (GNNs) due to their essential capacity for relational modeling. However, previous work often used a static graph structure of time-series variables for modeling MTS failing to capture their ever-changing correlations over time. To this end, a fully-connected supra-graph connecting any two variables at any two timestamps is adaptively learned to capture the high-resolution variable dependencies via an efficient graph convolutional network. Specifically, we construct the Edge-Varying Fourier Graph Networks (EV-FGN) equipped with Fourier Graph Shift Operator (FGSO) which efficiently performs graph convolution in the frequency domain. As a result, a high-efficiency scale-free parameter learning scheme is derived for MTS analysis and forecasting according to the convolution theorem. Extensive experiments show that EV-FGN outperforms state-of-the-art methods on seven real-world MTS datasets.