Forecasting The PM2.5 Concentrations In Urban Area

The air quality deterioration is gradually threatening our daily life, especially in the developing countries like China and India, and the air pollution issues have gained significant amount of attentions in recent years. It is critically important to predict the future air quality in advance, since it will be beneficial for people to make decisions. However, to the best of my knowledge, this paper is the first attempt to forecast the PM2.5 concentrations estimation in fine-granularity based on dense depolyment.The paper proposes a forecasting model by exploiting the local spatial similiarities and the global low rank feature of the spatial-temporal data matrix. The proposed model combines the traditional time series analysis model, includeing the linear forecasting model and the nonlinear regression model with the local similarities and the global low rank property of the spatial-temporal matrix by introducing a spatial and a global constraint matrix. Then the paper uses the real world PM2.5measurement data collected by 232 monitoring devices during six months in Beijing to validate the e?ectiveness of the proposed model, and the experiment results indicate that the model is capable of providing future estimation with acceptable prediction error and moderately outperforms the baseline time series forcasting model.