Abstract:Based on principal components analysis (PCA),the BP (Back Propagation) neural ne twork forecast method is introduced in air pollution prediction and the SO2 co ncentration prediction model is established. The results indicate that by applyi ng the principal component analysis in the data pre processing and taking the principal components of primitive predictor as the input of neural network, it can reduce the dimension of data, eliminate the correlation between the samples, and largel y speed up the convergence rate. The verification of forecast model shows that t he absolute error between the forecasts and the real value is 0.0098, and the co rrelation coefficient between them reaches 0.885. The PCA BP model has a fit ac curacy better than the common BP model.