Air Pollution Index Prediction Study Based On BP Neural Network Model And ARMA-BP Combined Model

Due to the rapid development of social economy,the problem of air pollutionin cities has become increasingly serious.This has become a global issue that directly threatens the sustainable development of the national economy and the physical and mental health of the people.It is of great significance to strengthen the prediction and research of atmospheric pollution.API is a quantitative scale that reflects and evaluates the quality of urban air quality,it has obvious advantages to study air pollution index.In this study,the daily average concentration of ambient air quality between Jan 1st,2014 to Dec 17th,2017 of Huai'an city is adopted and analyzed.Using R to calculate API according to the calculation method.In this paper,the daily air pollution index of spring and autumn is discussed.Firstly,the BP neural network model is used to forecast the two seasons'daily air pollution index respectively,the number of input layer,output layer and hidden layer is determined in the light of actual conditions,and the optimal initial weight,learning rate and dynamic coefficient are determined by repeated attempts.According to the test of forecasting the testing samples,the forecast value and the actual value are compared and analyzed.Secondly,the Hybrid Autoregressive moving Average(arma)and BP neural network model is adopted to forecast the two seasons'daily air pollution index respectively,the linear part of air pollution index is fitted by ARMA model,and the prediction residuals of ARMA model are predicted by BP neural network model,the forecast value and the actual value are compared and analyzed according to the test of forecasting the testing samples.Finally,by comparing the predictions of the two prediction models,on the basis of the average relative error of the test,the prediction ability and the generalization ability are also compared and analyzed.The results show that both models have a certain predictive effects,but the prediction ability of ARMA-BP neural network combined forecasting model is better than that of BP neural network model,and it is best to choose ARMA-BP Neural Network combined forecasting model for time series prediction of Air Pollution index.Because the ARMA-BP model contains both linear and nonlinear laws,the prediction accuracy is higher than using one single model,and the prediction result is more ideal.