Study Of Air Quality Based On Bayesian Networks In Dalian

With the rapid development of industrial technology and social economy in China,the problem of air pollution is becoming more and more prominent,meanwhile,it arouses the public attention.So the evaluation and prediction of ambient air quality is particularly important,which helps air quality warning early,giving public life guidance,reducing the impact on human health and economic losses.BN technology is introduced into the research of air quality in Dalian,exploratory research is developed.Bayesian network has advantages such as easy to understand,clear semantics and strong reason ability.In it causal relations and their probabilistic relations among variables can be explained effectively,and interaction between various factors on target variables can be considered.Aiming at the complexity and uncertainty of the evaluation and prediction of air quality.Through systematic theoretical and experimental analysis,the model is constructed.The study determine the factors affecting the air quality of the environment:SO2?NO2?CO?O3?PM2.5.PM10 and target factor--the air quality index(AQI),using the daily data of Dalian city in 2014-2016,after the noise reduction processing,calculating the mutual information between each factor and the objective factor,the variables that have significant correlation with the target variables are screened.Based on analyzing the characteristics of the sample data,we determine the BN structure learning algorithm and parameter learning algorithms:K2 algorithm and MAP algorithm,respectively.The model is constructed,and the results are as follows:1)the accuracy of BN model for the whole year and four seasons were higher than 85%,which the model validity was confirmed;2)the accuracy of the whole year and four seasons were 89.09%,89.29%,92.86%,88.89%,85.19%,respectively,higher than that of the fuzzy comprehensive evaluation method(FCEM);3)improved BN results using nonparametric regression algorithm and compare with the real value,the error value was within ±8,the average absolute error value of the forecast period was 2.49;4)for importance analysis:through analyzing the probability distribution of pollutants in the year and the four seasons,the primary pollutants were PM2.5.03?PM10;PM2.5?O3?CO;O3?PM2.5?PM10;PM2.5?O3?PM10;PM2.5?PM10?SO2 respectively,in the analysis of the probability distribution In air quality,it is concluded that the prediction of the ambient air quality is basically c-onsistent with the actual situation under the discrete grade of different pollutants.Finally,in the face of the environmental pollution state in Dalian,some suggestions are put forward to provide theoretical support for the environmental quality management.which is of great significance to the environmental quality management in Dalian.