Hybrid Prediction Model of Air Pollutant Concentration for PM2.5 and PM10

To alleviate the negative effects of air pollution, this paper explores a mixed prediction model pollutant concentration based on machine learning method. Firstly, in order to improve performance sparrow search algorithm least square support vector (SSA-LSSVM), reverse strategy-lens principle is introduced, and better solution obtained by optimizing current at same time. Secondly, according nonlinear non-stationary characteristics time series data PM2.5 PM10, variational mode decomposition (VMD) method used decompose original obtain appropriate K value. Finally, experimental verification an empirical analysis are carried out. In experiment 1, we verified good University California Irvine Machine Learning Repository (UCI) datasets. 2, predicted different cities Beijing–Tianjin–Hebei region periods, five error results compared them with six other algorithms. The show that has robustness expected can be under conditions.