The Research Of Clustering,Modeling And Prediction Of Air Quality Index In Some Cities Of China

The quality of air is closely related to people's daily life and physical health.This paper collects and arranges the air quality index(AQI)of 31 major cities in China over the past four years through the network and other methods.And uses them to cluster analysis on the air quality of these cities.Then,taking Urumqi as the representative,based on its AQI and related meteorological data,we establish the time series model,multiple linear regression model and BP neural network model,and we make the prediction and evaluation,and we obtain some meaningful conclusions.The main work of this paper follows:First,we calculate the quarterly seasonal index and monthly seasonal index of AQI in 31 major cities in China from October,2013 to October,2017,and make the descriptive statistical analysis of them,and then we use the hierarchical clustering and K-Means clustering methods to make the cluster analysis according to the quarterly average index and the monthly average index of the 31 cities.On this basis,the improved K-Means clustering method: PAM clustering method,is used to cluster the quarterly average index of 31 cities AQI,and the clustering result shows that the accuracy of the classification is 97.01%.Secondly,through the comprehensive comparative analysis,this paper uses Urumqi as the representative of the study,taking the data of Urumqi from October 28,2013 to October 31,2017 as the training set,taking the data from November 1,2017 to December 31,2017 as the prediction set,and using the training set to build mode of the single variable AQI and constructing the ARMA(2,2)model of the time series.On the basis of other scholars' research,from the point of view of AQI and meteorological factors,we collect related meteorological data that may affect AQI,and combine the training set to construct multiple linear regression model and BP neural network model.Finally,according to the prediction set,three models of ARMA(2,2),multiple linear regression and BP neural network are used to give the predictive value of AQI,and then the corresponding normalized mean square error(NMSE)of the three models is calculated,and the optimal model is selected by comparison.The results of this study show that the BP neural network model is relatively accurate and reliable for the prediction of AQI.