Study On PM₁₀-visibility-humidity Correlation Based On Machine Learning In Cloud Environment

With the rapid development of social economy,the pollution of the social environment was increasing,the fog haze caused by air pollutants was more and more frequent,and the decrease of visibility gradually brings some trouble to people's daily life.Therefore,it is necessary to analyze the correlation between aerosol PM₁₀,visibility and humidity.In the past of study,the correlation between PM₁₀ and visibility ?humidity was qualitatively based on mathematical statistics.However,with the accumulation of data in the meteorological field,the traditional linear regression method based on mathematical statistics can not be better to deal with meteorological relevance issues.The emergence of cloud computing to the meteorological data research into a new vitality,it can use very short period of time to calculate a large number of data,while part of the machine learning algorithm is also a good application in the field of meteorology.The main content of this paper is to study the correlation between PM₁₀ and visibility and humidity using the machine learning algorithm model in the cloud environment,which has certain application significance.And at present,there is no study on the correlation between the aerosol PM₁₀ and the visibility and humidity,it is also provides a basis and a new direction for the subsequent research.The main work of this paper is as follows:?1?In the cloud environment to establish a correlation analysis of the experimental architecture platform,the framework includes computing engine and storage,machine learning algorithm model,predictive evaluation and the results of the show.?2?Based on the idea of multiple linear regression,random forest and logistic regression algorithm,we design DMLR?Distributed Multiple Linear Regression?model,DRF?Distributed Random Forests?model and DLR?Distributed Logistic Regression?model for PM₁₀-Visibility-humidity correlation study in cloud environment.?3?The Realization of PM₁₀-Visibility-Humidity Correlation Research in Cloud Environment.Firstly,the use of cloud platform at the bottom of the computing engine services on the PM₁₀,visibility and humidity of the original data table for import,then through use the database operating language on the import of the humidity indicator table and visibility index table to split and connect for generate spare data sets.;And then use the DMLR model,DRF model and DLR model to process each sample data set,and then input the corresponding sample test set to get the forecast result;Finally,analyze the forecasting result and compare with the corresponding evaluation standard.The feasibility of the model and the correlation between PM₁₀,visibility and humidity.The experimental results show that the smaller the visibility value is in the same relative humidity range,the atmospheric aerosol concentration is too large,the larger the relative humidity value is in the same visibility range,and the atmospheric aerosol concentration is low.And the humidity value between 40%-90%,the visibility value between 8km-19 km PM₁₀ value of the best results and the correlation coefficient is high.Based on the idea of two categories,DLR model is better than DRF model for the data set studied in this paper.In terms of performance comparison,the DLR model is better than DMLR.