Research On The Influencing Factors And Prediction Of Urban Air Quality Based On Machine Learning

Jinan is located in the north of China.As the capital city of Shandong Province,its economic development level has been far higher than the average domestic development level.However,in recent years,due to the terrain of Jinan and the rapid development of industrial transportation and other industries,the air quality of Jinan is not satisfactory.The air pollution in Jinan has always affected and troubled the daily behavior and health of residents.At the same time,the problem has also been widely valued by all walks of life.Through collecting the data of air pollutants and meteorological indicators in Jinan in recent years,this thesis studies the air quality impact factors and future AQI prediction in Jinan,as follows:First of all,relevant data of air quality in Jinan are obtained through crawler technology,and the data are preprocessed,described and visualized to preliminarily explore the change law of air quality in Jinan;Secondly,the influence factors of air quality are studied,and the relationship between meteorological factors and six pollutants,and the relationship between meteorological factors and air quality index AQI are explored by establishing multiple regression models;Then,based on ARIMA model,random forest model,XGBoost model and support vector machine,the AQI of Jinan is predicted and Goodness of Fit,MSE and MAE are used as evaluation criteria to evaluate and compare the prediction effect of each prediction model on AQI series;Then,aiming at the above stochastic forest model and XGBoost model with better prediction effect,the residual error of the stochastic forest model and XGBoost model on the AQI prediction results is corrected by using support vector machine,and two combined prediction models,RF-SVR and XGBoost-SVR,are proposed and constructed.Finally,the prediction effect of all models on the AQI in Jinan is comprehensively compared and analyzed through the evaluation criteria.The main research results obtained in this thesis are reflected in the air quality influencing factors and air quality index prediction in Jinan.First,among the six pollutants,PM2.5 and PM10 have the greatest impact on AQI.Among the meteorological factors,the order of impact on Jinan air quality index AQI is: temperature>pressure>wind level>precipitation>wind direction>humidity;Second,the established random forest model and RF-SVR combined prediction model have the best prediction effect on the AQI of Jinan,and the evaluation standard is Goodness of Fit,MSE and MAE reach 0.998556,1.119044,0.246292 and 0.998572,1.107289,0.243125,respectively,of which the RF-SVR combined prediction model is more ideal than the single random forest model.This study has certain reference significance for the relevant departments in Jinan to monitor the air quality status in a timely and effective manner and formulate policies to improve air pollution.