Model For The Analysis And Prediction Of Air Quality Based On Spatial-temporal Correlation

With the increase in industrialization,air pollution has increasingly become a serious social problem.Pollution of the air can result in smog and acid rain,and long-term exposure to environmental pollution can lead to many illnesses.Pollution control has become an important element of government work in recent years in order to ensure the green,healthy,and sustainable development of society.By understanding air quality changes,pollution control can be facilitated;therefore,further research on air quality prediction algorithms has important application value and practical significance.Using a classification of air quality influencing factors,data time series,and spatial distribution characteristics,this paper proposes a data selection method,adds a feature extraction mechanism to the LSTM model,and conducts algorithm research to enhance the generalization capability of the air quality prediction model.The main research contents are as follows:(1)Using three methods: Assess air quality from the perspective of its temporal variation characteristics,spatial variation characteristics,pollution source classification,pollution control measures,and regional economic development factors,etc.Analyze data characteristics using kernel density estimation,spatial auto-correlation method,and random forest.On this basis,a summary of five site classification methods is provided.(2)A data selection method,PSS(Pearson Based Station Selection)has been proposed to compare the predictive effects of different classification methods: first,similar stations are categorized as a set and the Pearson correlation coefficient is used to evaluate the correlation of changes in data based on the required number of observations.A higher level of site information is added to the model.As a result of this method,it is possible to better utilize the air quality monitoring sites and the changing characteristics of data when selecting data,thereby increasing the efficiency of data input.PSS under the same model produces predictions with smaller error values than other methods.(3)STF-LSTM(Spatial Temporal Feature based LSTM Model)is an air quality prediction model that selects data based on data and site spatiotemporal feature correlation and adds a mechanism for feature extraction.Convolution operation is used to extract small-scale features from the data based on PSS site selection,while the attention mechanism is used to extract time-series features.The experimental results indicate that the addition of a feature extraction and attention mechanism improves the stability of the prediction model,as well as its ability to generalize to process input data.