Air Quality Multi-mode Ensemble Prediction Method Research And Application Based On LSTM Neural Network

The explosion of the air pollutant concentration has become one of the hottest issues in the society.The accurate prediction of air quality is of significance to practical activities of human health and disease prevention and so on.From the perspective of data science,this paper introduces the prediction tool of LSTM neural network,and proposes an optimized VMD-ELSTM-GS model for the prediction of air quality in Lanzhou under the "decomposition-integration" paradigm,which is expected to improve the prediction accuracy of target pollutants.At the same time,we developed a static web page for the release of prediction results.This article includes mainly:Firstly,the process of data exploration and preprocessing about meteorological data and air quality monitoring data is described,and we introduces LSTM neural network to the practical task of air quality prediction.Execute data preprocessing based on the theories of missing value processing,outlier processing,and data reshape to improve data quality;Perform feature selection based on correlation theory;Construct LSTM neural network models that adapt to time series prediction to forecast air pollutants with multi-indicators.Secondly,in order to adapt to air quality data's characteristics of multiple indicators,high fluctuation,non-stationary,noise,and improve the prediction accuracy,a VMD-ELSTM-GS prediction model was proposed.A type of hybrid,multi-mode integrated prediction model is constructed for air quality prediction research based on the theory of optimization algorithms,signal decomposition,deep learning and regularization.This model was used in the empirical process of the 2016-2017 meteorological and air quality data sets about Lanzhou,and was compared with the experimental group methods in the indicators of root mean square error(RMSE)and absolute value error(MAE),which shows the superiority of the model empirically.Finally,we develop a static webpage to publish the prediction results of air quality index and pollutant concentration,and introduce its release process briefly.The goal of timely and convenient release of air quality prediction results was achieved at the technical level.Explore ways for practical application of semi-structural data.