Research And Application Of Air Quality Forecast Based On Deep Learning

With the development of industrialization,it has brought great material achievements to people.But at the same time,the air pollution issues have become more and more seriously.It not only destroys the environment of earth,but also impacts the human health and development.Based on this,people pay more attention to the air quality situation.Since the 1970 s,the methods of air quality prediction are emerging in an endless stream.However,the methods’ accuracies are not high and they can only be used in limited scenarios.With the advancement of science and technology and the information disclosure,the development of new technology has provided new ideas for the field of air quality predicting and improving the accuracy of predicting becomes practical and feasible,which has become one of the hot research topics.In this thesis,the air quality prediction as the point of penetration,with data processing and analysis,the extraction of data time dimension characteristics,the determination of data spatial correlation and the acquisition of data spatial dimension characteristics as the main line.From these,it carried out the relevant theoretical foundation research and application work.Firstly,the research analyzed the current research status of air quality prediction methods at home and abroad and summarized the shortcomings of the current methods.Secondly,the relevant basic theoretical knowledge involved in the air quality prediction methods are studied,which lay a theoretical foundation for the subsequent prediction method.Thirdly,based on the analysis of data correlation,a prediction method based on the time dimension of stacked denoising autoencoder(SDAE)combined with GRU neural network,a spatial region selection method,and a graph convolutional neural network(GCN)based on the space-time dimension are proposed.The prediction method combined with GRU neural network,and experiments on the proposed method with real data,the experimental results prove the feasibility and advantages of the method.Finally,combined with actual application scenarios,an air quality prediction system is developed and implemented,and the paper is proposed.The air quality prediction method used in the actual system,and the system has been tested its functions.The results proved the feasibility and practicability of the system.Through the above research and experimental demonstration,the air quality prediction system based on deep learning has proved the practical feasibility of the research method,which can provide new ideas for the development of prediction methods,and has practical research significance.