Classification And Prediction Of Air Quality Based On Machine Learning

Air pollution refers to the phenomenon that human beings carry out production and life in the natural environment and bring some pollutants into the atmosphere because of some inappropriate behavior.When the concentration of material reaches a certain value,it is harmful to human health and the natural environment.The atmospheric environment is complex and changeable and has dynamic uncertainty.There are many factors that cause air pollution,such as PM_2.5,SO₂ and O₃,which are more than certain concentration of air pollutants,as well as precipitation,wind direction,humidity and so on.These factors have a strong nonlinear relationship with the air quality within a period of time.Therefore,the more accurate air quality forecast can help people to take effective measures and arrange the trip plan reasonably.It can help the prevention and control of air pollution and the construction of urban environmental planning,reduce unnecessary losses,and have important guiding significance for people's production and life.In recent years,artificial intelligence has made new progress and breakthroughs.Machine learning has played a very important role.At present,researchers in various fields of the whole society are using machine learning to solve difficult problems,making machine learning a hot processing method.In this thesis,the intelligent optimization algorithm and the support vector machine and neural network in machine learning are used to study the air pollutant monitoring data from 2014 to 2017 in Taiyuan.The main contents are as follows:?1?The simulated annealing algorithm?SA?and particle swarm optimization?PSO?are combined and improved to optimize support vector machine?SVM?for parameter optimization,and the partial least square method?PLS?is used to analyze the interaction among pollutant factors and propose a new air quality evaluation model.Experimental results show that compared with PSO-SVM and SVM,the improved SAPSO-SVM has shorter running time,higher classification accuracy and better evaluation performance.?2?the improved particle swarm optimization?IPSO?and genetic algorithm?GA?and support vector machine?SVM?are combined to build a new model to predict the air quality index?AQI?.Experimental results show that GA-SVM is superior to IPSO-SVM and SVM in terms of prediction accuracy,error rate and reliability.?3?using the fruit fly optimization algorithm?FOA?to optimize the grey neural network?GMNN?,optimize the network parameters and predict the future air quality in Taiyuan.Experimental results show that FOA-GMNN is superior to GMNN in prediction accuracy,error rate and reliability,showing the good performance of the algorithm.The three prediction models proposed in this thesis,the improved SAPSO-SVM,GA-SVM and FOA-GMNN,are used for the classification and prediction of air quality.It provides a new idea for air quality evaluation and provides a scientific and rational theoretical basis and a new prediction method for the prevention and control of air pollution.