Multivariate Time-series Forecasting Using GNN

Abstract of Spectral Temporal Graph Neural Network for multivariate time-series forecasting (arXiv:2103.07719):

Multivariate time-series forecasting plays a crucial rolein many real-world ap-plications. It is a challenging problem as one needs to consider both intra-seriestemporal correlations and inter-series correlations simultaneously. Recently, there have been multiple works trying to capture both correlations, but most, if not allof them only capture temporal correlations in the time domain and resort to pre-defined priors as inter-series relationships. In this paper, we propose Spectral Temporal Graph Neural Network (StemGNN)to further improve the accuracy of multivariate time-series forecasting. StemGNNcaptures inter-series correlations and temporal dependencies jointlyin thespec-tral domain. It combines Graph Fourier Transform (GFT) which models inter-series correlations and Discrete Fourier Transform(DFT) which models temporal dependencies in an end-to-end framework. After passing throughGFT and DFT, the spectral representations hold clear patterns and can be pre-dicted effectively by convolution and sequential learning modules. Moreover, StemGNN learns inter-series correlations automatically from the data without using pre-defined priors. We conduct extensive experimentson ten real-worlddatasets to demonstrate the effectiveness of StemGNN. Codeis available at https://github.com/microsoft/StemGNN/