RNN with Particle Flow for Probabilistic Spatio-temporal Forecasting

Soumyasundar Pal, Liheng Ma, Yingxue Zhang, Mark Coates
Proceedings of the 38th International Conference on Machine Learning, PMLR 139:8336-8348, 2021.

Abstract

Spatio-temporal forecasting has numerous applications in analyzing wireless, traffic, and financial networks. Many classical statistical models often fall short in handling the complexity and high non-linearity present in time-series data. Recent advances in deep learning allow for better modelling of spatial and temporal dependencies. While most of these models focus on obtaining accurate point forecasts, they do not characterize the prediction uncertainty. In this work, we consider the time-series data as a random realization from a nonlinear state-space model and target Bayesian inference of the hidden states for probabilistic forecasting. We use particle flow as the tool for approximating the posterior distribution of the states, as it is shown to be highly effective in complex, high-dimensional settings. Thorough experimentation on several real world time-series datasets demonstrates that our approach provides better characterization of uncertainty while maintaining comparable accuracy to the state-of-the-art point forecasting methods.

Cite this Paper

BibTeX
@InProceedings{pmlr-v139-pal21b, title = {RNN with Particle Flow for Probabilistic Spatio-temporal Forecasting}, author = {Pal, Soumyasundar and Ma, Liheng and Zhang, Yingxue and Coates, Mark}, booktitle = {Proceedings of the 38th International Conference on Machine Learning}, pages = {8336--8348}, year = {2021}, editor = {Meila, Marina and Zhang, Tong}, volume = {139}, series = {Proceedings of Machine Learning Research}, month = {18--24 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v139/pal21b/pal21b.pdf}, url = {https://proceedings.mlr.press/v139/pal21b.html}, abstract = {Spatio-temporal forecasting has numerous applications in analyzing wireless, traffic, and financial networks. Many classical statistical models often fall short in handling the complexity and high non-linearity present in time-series data. Recent advances in deep learning allow for better modelling of spatial and temporal dependencies. While most of these models focus on obtaining accurate point forecasts, they do not characterize the prediction uncertainty. In this work, we consider the time-series data as a random realization from a nonlinear state-space model and target Bayesian inference of the hidden states for probabilistic forecasting. We use particle flow as the tool for approximating the posterior distribution of the states, as it is shown to be highly effective in complex, high-dimensional settings. Thorough experimentation on several real world time-series datasets demonstrates that our approach provides better characterization of uncertainty while maintaining comparable accuracy to the state-of-the-art point forecasting methods.} }
Endnote
%0 Conference Paper %T RNN with Particle Flow for Probabilistic Spatio-temporal Forecasting %A Soumyasundar Pal %A Liheng Ma %A Yingxue Zhang %A Mark Coates %B Proceedings of the 38th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2021 %E Marina Meila %E Tong Zhang %F pmlr-v139-pal21b %I PMLR %P 8336--8348 %U https://proceedings.mlr.press/v139/pal21b.html %V 139 %X Spatio-temporal forecasting has numerous applications in analyzing wireless, traffic, and financial networks. Many classical statistical models often fall short in handling the complexity and high non-linearity present in time-series data. Recent advances in deep learning allow for better modelling of spatial and temporal dependencies. While most of these models focus on obtaining accurate point forecasts, they do not characterize the prediction uncertainty. In this work, we consider the time-series data as a random realization from a nonlinear state-space model and target Bayesian inference of the hidden states for probabilistic forecasting. We use particle flow as the tool for approximating the posterior distribution of the states, as it is shown to be highly effective in complex, high-dimensional settings. Thorough experimentation on several real world time-series datasets demonstrates that our approach provides better characterization of uncertainty while maintaining comparable accuracy to the state-of-the-art point forecasting methods.
APA
Pal, S., Ma, L., Zhang, Y. & Coates, M.. (2021). RNN with Particle Flow for Probabilistic Spatio-temporal Forecasting. Proceedings of the 38th International Conference on Machine Learning, in Proceedings of Machine Learning Research 139:8336-8348 Available from https://proceedings.mlr.press/v139/pal21b.html.

Related Material