Air Quality Index Prediction Based On Fuzzy Information Granulation And ARIMA-SVR Combined Model

In recent years,with the rapid development of China’s industry and economy,China’s air pollution problem has become more and more serious,not only affecting the safety of people’s travel,but also seriously harming people’s health.Therefore,in order to effectively improve air quality,and provide some reference suggestions for travel safety,it is necessary to monitor and forecast air quality in a timely manner.Air Quality Index(AQI)is an indicator that reflects air quality.The larger the value,the more significant the air pollution.Therefore,predicting AQI is vital to solve the air pollution problem.At present,commonly used AQI prediction methods include ARIMA model,SVR model,neural network models,etc.,which are usually point predictions.With the increasing number of samples and the increasing dimension of data,the calculation process of point prediction becomes more and more complicated,and the amount of calculation also increases significantly.Due to the complex and volatile nature of the atmospheric environment,it is impossible to accurately predict the point value of AQI most of the time,so it is very important to be able to predict the interval of AQI.The fuzzy information granulation model can simplify the high-dimensional large sample data,reduce the amount of calculation,while retaining the valuable information in the sample,and can use the simplified data to perform interval prediction on the original time series data.Although interval prediction can not give researchers accurate point predictions,it can give researchers the range and trend of data changes in the future.AQI has both the overall linear characteristics of the time series in fluctuations,but also the uncertainty of various factors Qualitatively,in order to improve the prediction accuracy of AQI,this paper uses a ARIMA-SVR combined model,a time series prediction model with both linear and nonlinear composite features.This model uses the ARIMA model for linear prediction,and uses the SVR model for non-linear prediction of the ARIMA model prediction residuals,and adds up to obtain the prediction result of the combined prediction model.In this paper,the fuzzy information granulation model and the ARIMA-SVR combination model are combined to construct an AQI interval prediction model to conduct three-day prediction of the change range and trend of AQI.Based on the AQI daily data from January 1,2017 to October 28,2019 in the five cities(Beijing,Shanghai,Guangzhou,Jinan and Zhengzhou),the fuzzy information granulation model was first used to granulate the AQI data of the five cities.Then,the ARIMA model,SVR model and ARIMA-SVR combined model were used to make regression predictions on the AQI fuzzy particles,and the prediction results of the AQI interval of the five cities were obtained.The accuracy rate,accuracy and reliability of the trend prediction,and model stability of the three models were analyzed and compared,the results show that the combined model exerts the respective advantages of the two models.Compared with the single forecasting model,the accuracy and reliability of the trend prediction and the range prediction are improved in the combined model,which also has higher stability.Finally,the combined model is applied to the air quality prediction in the five cities.The corresponding results obtained through this model not only can provide guidance for individuals living in the city,but also can provide scientific basis and reference for the prevention and control of air pollution.