Forecasting of bioaerosol concentration by a Back Propagation neural network model

Abstract Bioaerosol in the atmosphere plays a very important role in environment and public health. To forecast the bioaerosol concentration, the correlation between bioaerosol concentration and meteorological factors was discussed, and a Back Propagation (BP) neural network with Principal Component Analysis (PCA) method was utilized in this study. The proposed method works in three steps. The first step is to compute the correlation between bioaerosol concentration and meteorological factors, which consists of analyzing correlation and selecting meteorological factors applied to the study of forecast model. The second step is to use PCA analysis to reduce the dimensions of meteorological dataset. The third step is to use BP neural network, setting up, training BP neural network and proving the feasibility of forecast model included. The results of our model in forecasting bioaerosol concentration show 10.55% of average relative error, 2.80 pieces/L (pcs/L) of average absolute error, and 84.01 grade of forecast accuracy, providing a promising model for the forecasting of bioaerosol concentration.