MSSTN: Multi-Scale Spatial Temporal Network for Air Pollution Prediction.

Zhiyuan Wu,Yue Wang,Lin Zhang
DOI: https://doi.org/10.1109/bigdata47090.2019.9005574
2019-01-01
Abstract:Air pollution has become an important factor constraining city development and threatening public health in recent years. Air pollution prediction has been considered as the key part for the early warning of pollution event. Considering the multi-scale nature of geo-sensory data such as air pollution signal, in this paper we adopt a multi-level graph data structure for better utilization of multi-scale spatio-temporal information. We further present a novel deep convolutional neural network model, named Multi-Scale Spatial Temporal Network (MSSTN), for the learning task on this data structure. The MSSTN is specially designed to better discover multi-scale spatial temporal patterns and their high-level interactions, by explicitly using multi-scale neural network structure in both spatial and temporal component. We conduct extensive experiments and ablation studies on Urban Air Pollution Datasets in North China, where the MSSTN can make hourly PM2.5 concentration predictions jointly for a number of cities. And our results shows an outstanding prediction accuracy as well as high computational efficiency compared to existing works.
What problem does this paper attempt to address?