Application Research Of Time Series And Regression Hybrid Analysis In Air Quality Forecast

Ambient air pollution is the biggest environmental risk faced by global public health.Whether in developed or developing countries,air pollutants have a very serious impact on the health of the general public.In recent years,air pollution in Xi ’an has become more and more serious,with frequent occurrence of haze phenomenon.Problems such as environmental air quality analysis and evaluation and pollution control have once again attracted public attention.It is an effective way to improve air quality to forecast and analyze air quality and implement timely management measures,which has important and realistic significance for improving social living environment.In this paper,deep learning and ensemble learning related technologies are introduced into air quality prediction,which has certain guiding significance for the prediction and prevention of environmental air pollution.(1)Time series prediction of air pollutants based on 1D-CNN-LSTM hybrid neural network.Based on the study of the existing time series prediction and analysis technology,this paper proposed to use 1D-CNN layer to extract the sub-series model of time series,and to build a 1DCNN-LSTM hybrid neural network model,optimize the prediction effect of LSTM network,so as to predict the value of air pollution components in Xi ’an area.Through the simulation experiment,the feasibility and effectiveness of the hybrid neural network model are verified,and the accuracy of the time series prediction of air pollution components is improved.(2)Time series anomaly detection of air pollutants based on Seq2 Seq deep autoencoder.On the basis of studying the principle of the existing time series anomaly detection technology,this paper proposes an unsupervised time series anomaly detection method.Based on Seq2 Seq model,this method uses Bi-LSTM deep autoencoder to reconstruct the sequence to find abnormal sequence.The discovery of abnormal time series depends on the reconstruction effect of the model to the original sequence.This method can be used to detect and provide abnormal sample data,and can better mine the abnormal sequence patterns in the time series data.The feasibility of the proposed anomaly detection method is verified by the simulation experiment of air pollution time series data.(3)Research on air quality regression prediction based on bayesian optimization gradient boosting algorithm.On the basis of studying the existing air quality index prediction methods,this paper constructs gradient boosting model to make regression prediction of air quality index.And bayesian optimization algorithm is used to optimize the hyperparameters in the gradient boosting model,so as to establish the prediction model with the best combination of hyperparameters.The validity of the proposed method is verified by the experiment of air quality data from 2014-1-20 to 2020-5-11 in Xi ’an area.