Deep Multi-task Learning for Air Quality Prediction

Bin Wang, Zheng Yan, Jie Lu, Guangquan Zhang, Tianrui Li

Abstract: Predicting the concentration of air pollution particles has been an important task of urban computing. Accurately measuring and estimating makes the citizen and governments can behave with suitable decisions. In order to predict the concentration of several air pollutants at multiple monitoring stations throughout the city region, we proposed a novel deep multi-task learning framework based on residual Gated Recurrent Unit (GRU). The experimental results on the real world data from London region substantiate that the proposed deep model has manifest superiority than shallow models and outperforms 9 baselines.