Spatio-Temporal Neural Networks for Space-Time Series Forecasting and Relations Discovery

TL;DR

This paper presents a recurrent neural network model for forecasting space-time series that captures spatial and temporal dependencies, and can also discover underlying spatial relations across various application domains.

Contribution

The paper introduces a structured latent dynamical model for space-time series forecasting that also uncovers spatial relations, with multiple variants and comprehensive evaluations.

Findings

Outperforms state-of-the-art baselines in forecasting accuracy.

Effectively extracts meaningful spatial relations from data.

Applicable across epidemiology, geo-statistics, and traffic prediction.

Abstract

We introduce a dynamical spatio-temporal model formalized as a recurrent neural network for forecasting time series of spatial processes, i.e. series of observations sharing temporal and spatial dependencies. The model learns these dependencies through a structured latent dynamical component, while a decoder predicts the observations from the latent representations. We consider several variants of this model, corresponding to different prior hypothesis about the spatial relations between the series. The model is evaluated and compared to state-of-the-art baselines, on a variety of forecasting problems representative of different application areas: epidemiology, geo-spatial statistics and car-traffic prediction. Besides these evaluations, we also describe experiments showing the ability of this approach to extract relevant spatial relations.