A Hybrid Model For PM2.5 Forecasting Of Cities Based On Ensemble Empirical Mode Decomposition And A Random Forest

In recent years,air pollution has become a serious major issue affecting the quality of people's life and travel,especially PM2.5 pollution.The annual average PM2.5 concentrations of some large and medium-sized cities in our country is higher than the world health organization environment air quality guideline values 10ug/m3,cardiovascular and respiratory diseases increased year by year.As the main pollutants of forming fog haze,PM2.5 has become the primary target of urban air pollution.So it is important to research the trend of PM2.5 and establish a reasonable and accurate forecasting model of PM2.5.It not only can remind people to take effective measures to protect the healthy body,but also can regulate some social activities.It also has a practical significance for early warning the harm of severe environmental pollution.In order to build a reasonable PM2.5 prediction model,the overall experience mode decomposition method of PM2.5 in Beijing time series,study the volatile regularity of PM2.5 in Beijing and its periodic changes.In this paper,we take Beijing as an example.The ensemble empirical mode decomposition(EEMD)method is used to decompose original PM2.5 data to study the fluctuation rule of PM2.5 in Beijing and their periodic changes.Firstly,we analyze the trend of changes of PM2.5,volatile and periodic changes of PM2.5.Secondly,we use principal component regression(PCR)analysis method,support vector machine regression(SVR)method,ARIMA method and random forest regression method to predict the PM2.5.Finally,a hybrid model for PM2.5 forecasting of cities based on ensemble empirical mode decomposition and a random forest(EEMD-RF)is proposed.In this part,the EEMD method is used to decompose original PM2.5 data into several intrinsic mode function(IMFs),and the RF method is utilized to predict each IMF.Then all the prediction method are compared in order to find a best model to give some recommendations to the environmental protection department in the process of governance PM2.5.