Spectral Temporal Graph Neural Network for Multivariate Time-series Forecasting. Cao, D., Wang, Y., Duan, J., Zhang, C., Zhu, X., Huang, C., Tong, Y., Xu, B., Bai, J., Tong, J., & Zhang, Q. 2021. cite arxiv:2103.07719Comment: Accepted by NeurIPS 2020. 20 pages, 7 figures![Spectral Temporal Graph Neural Network for Multivariate Time-series Forecasting [link]](https://bibbase.org/img/filetypes/link.svg "link")
Paper abstract bibtex Multivariate time-series forecasting plays a crucial role in many real-world applications. It is a challenging problem as one needs to consider both intra-series temporal correlations and inter-series correlations simultaneously. Recently, there have been multiple works trying to capture both correlations, but most, if not all of 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. StemGNN captures inter-series correlations and temporal dependencies jointly in the spectral 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 through GFT and DFT, the spectral representations hold clear patterns and can be predicted 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 experiments on ten real-world datasets to demonstrate the effectiveness of StemGNN. Code is available at https://github.com/microsoft/StemGNN/
@misc{cao2021spectral,
abstract = {Multivariate time-series forecasting plays a crucial role in many real-world
applications. It is a challenging problem as one needs to consider both
intra-series temporal correlations and inter-series correlations
simultaneously. Recently, there have been multiple works trying to capture both
correlations, but most, if not all of 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. StemGNN
captures inter-series correlations and temporal dependencies \textit{jointly}
in the \textit{spectral 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
through GFT and DFT, the spectral representations hold clear patterns and can
be predicted 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 experiments on ten
real-world datasets to demonstrate the effectiveness of StemGNN. Code is
available at https://github.com/microsoft/StemGNN/},
added-at = {2023-04-12T14:27:15.000+0200},
author = {Cao, Defu and Wang, Yujing and Duan, Juanyong and Zhang, Ce and Zhu, Xia and Huang, Conguri and Tong, Yunhai and Xu, Bixiong and Bai, Jing and Tong, Jie and Zhang, Qi},
biburl = {https://www.bibsonomy.org/bibtex/2e30d6b0586c9494400f9c2b72c8b5612/manli},
description = {Spectral Temporal Graph Neural Network for Multivariate Time-series Forecasting},
interhash = {78af5a11e4c62590f551cc7c88e35c4d},
intrahash = {e30d6b0586c9494400f9c2b72c8b5612},
keywords = {ma_sem2023},
note = {cite arxiv:2103.07719Comment: Accepted by NeurIPS 2020. 20 pages, 7 figures},
timestamp = {2023-04-12T14:27:15.000+0200},
title = {Spectral Temporal Graph Neural Network for Multivariate Time-series
Forecasting},
url = {http://arxiv.org/abs/2103.07719},
year = 2021
}
Downloads: 0
{"_id":"zpxXAyKn5Zg7nBwyg","bibbaseid":"cao-wang-duan-zhang-zhu-huang-tong-xu-etal-spectraltemporalgraphneuralnetworkformultivariatetimeseriesforecasting-2021","author_short":["Cao, D.","Wang, Y.","Duan, J.","Zhang, C.","Zhu, X.","Huang, C.","Tong, Y.","Xu, B.","Bai, J.","Tong, J.","Zhang, Q."],"bibdata":{"bibtype":"misc","type":"misc","abstract":"Multivariate time-series forecasting plays a crucial role in many real-world applications. It is a challenging problem as one needs to consider both intra-series temporal correlations and inter-series correlations simultaneously. Recently, there have been multiple works trying to capture both correlations, but most, if not all of 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. StemGNN captures inter-series correlations and temporal dependencies <i>jointly</i> in the <i>spectral domain</i>. 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 through GFT and DFT, the spectral representations hold clear patterns and can be predicted 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 experiments on ten real-world datasets to demonstrate the effectiveness of StemGNN. Code is available at https://github.com/microsoft/StemGNN/","added-at":"2023-04-12T14:27:15.000+0200","author":[{"propositions":[],"lastnames":["Cao"],"firstnames":["Defu"],"suffixes":[]},{"propositions":[],"lastnames":["Wang"],"firstnames":["Yujing"],"suffixes":[]},{"propositions":[],"lastnames":["Duan"],"firstnames":["Juanyong"],"suffixes":[]},{"propositions":[],"lastnames":["Zhang"],"firstnames":["Ce"],"suffixes":[]},{"propositions":[],"lastnames":["Zhu"],"firstnames":["Xia"],"suffixes":[]},{"propositions":[],"lastnames":["Huang"],"firstnames":["Conguri"],"suffixes":[]},{"propositions":[],"lastnames":["Tong"],"firstnames":["Yunhai"],"suffixes":[]},{"propositions":[],"lastnames":["Xu"],"firstnames":["Bixiong"],"suffixes":[]},{"propositions":[],"lastnames":["Bai"],"firstnames":["Jing"],"suffixes":[]},{"propositions":[],"lastnames":["Tong"],"firstnames":["Jie"],"suffixes":[]},{"propositions":[],"lastnames":["Zhang"],"firstnames":["Qi"],"suffixes":[]}],"biburl":"https://www.bibsonomy.org/bibtex/2e30d6b0586c9494400f9c2b72c8b5612/manli","description":"Spectral Temporal Graph Neural Network for Multivariate Time-series Forecasting","interhash":"78af5a11e4c62590f551cc7c88e35c4d","intrahash":"e30d6b0586c9494400f9c2b72c8b5612","keywords":"ma_sem2023","note":"cite arxiv:2103.07719Comment: Accepted by NeurIPS 2020. 20 pages, 7 figures","timestamp":"2023-04-12T14:27:15.000+0200","title":"Spectral Temporal Graph Neural Network for Multivariate Time-series Forecasting","url":"http://arxiv.org/abs/2103.07719","year":"2021","bibtex":"@misc{cao2021spectral,\n abstract = {Multivariate time-series forecasting plays a crucial role in many real-world\r\napplications. It is a challenging problem as one needs to consider both\r\nintra-series temporal correlations and inter-series correlations\r\nsimultaneously. Recently, there have been multiple works trying to capture both\r\ncorrelations, but most, if not all of them only capture temporal correlations\r\nin the time domain and resort to pre-defined priors as inter-series\r\nrelationships.\r\n In this paper, we propose Spectral Temporal Graph Neural Network (StemGNN) to\r\nfurther improve the accuracy of multivariate time-series forecasting. StemGNN\r\ncaptures inter-series correlations and temporal dependencies \\textit{jointly}\r\nin the \\textit{spectral domain}. It combines Graph Fourier Transform (GFT)\r\nwhich models inter-series correlations and Discrete Fourier Transform (DFT)\r\nwhich models temporal dependencies in an end-to-end framework. After passing\r\nthrough GFT and DFT, the spectral representations hold clear patterns and can\r\nbe predicted effectively by convolution and sequential learning modules.\r\nMoreover, StemGNN learns inter-series correlations automatically from the data\r\nwithout using pre-defined priors. We conduct extensive experiments on ten\r\nreal-world datasets to demonstrate the effectiveness of StemGNN. Code is\r\navailable at https://github.com/microsoft/StemGNN/},\n added-at = {2023-04-12T14:27:15.000+0200},\n author = {Cao, Defu and Wang, Yujing and Duan, Juanyong and Zhang, Ce and Zhu, Xia and Huang, Conguri and Tong, Yunhai and Xu, Bixiong and Bai, Jing and Tong, Jie and Zhang, Qi},\n biburl = {https://www.bibsonomy.org/bibtex/2e30d6b0586c9494400f9c2b72c8b5612/manli},\n description = {Spectral Temporal Graph Neural Network for Multivariate Time-series Forecasting},\n interhash = {78af5a11e4c62590f551cc7c88e35c4d},\n intrahash = {e30d6b0586c9494400f9c2b72c8b5612},\n keywords = {ma_sem2023},\n note = {cite arxiv:2103.07719Comment: Accepted by NeurIPS 2020. 20 pages, 7 figures},\n timestamp = {2023-04-12T14:27:15.000+0200},\n title = {Spectral Temporal Graph Neural Network for Multivariate Time-series\r\n Forecasting},\n url = {http://arxiv.org/abs/2103.07719},\n year = 2021\n}\n\n","author_short":["Cao, D.","Wang, Y.","Duan, J.","Zhang, C.","Zhu, X.","Huang, C.","Tong, Y.","Xu, B.","Bai, J.","Tong, J.","Zhang, Q."],"key":"cao2021spectral","id":"cao2021spectral","bibbaseid":"cao-wang-duan-zhang-zhu-huang-tong-xu-etal-spectraltemporalgraphneuralnetworkformultivariatetimeseriesforecasting-2021","role":"author","urls":{"Paper":"http://arxiv.org/abs/2103.07719"},"keyword":["ma_sem2023"],"metadata":{"authorlinks":{}},"html":""},"bibtype":"misc","biburl":"http://www.bibsonomy.org/bib/author/zhang?items=1000","dataSources":["6yXn8CtuzyEbCSr2m"],"keywords":["ma_sem2023"],"search_terms":["spectral","temporal","graph","neural","network","multivariate","time","series","forecasting","cao","wang","duan","zhang","zhu","huang","tong","xu","bai","tong","zhang"],"title":"Spectral Temporal Graph Neural Network for Multivariate Time-series Forecasting","year":2021}