Research On Environmental Monitoring Data Analysis And Forecasting Model

Air pollution affects our health and the environment.In 2013,the State Council issued"Action Plan on Prevention and Control of Air Pollution"(ten measure of atmosphere),as well as national key R&D projects in the environmental special policy development and financial support,all indicate the country's attention to environmental issues.This paper analyzes the environmental data and studies the prediction model,which consists of five sections.The first section explains the research background,the content and the research significance.With the public's concern about the air pollution index in the living area,the atmospheric monitoring data from the past only a few key monitoring points,developed to the regional grid layout,making the monitoring of time and space data increased rapidly.The traditional air quality prediction model is more complex because the data collection point is small,can only be closely combined with geographic weather and other external data,data is not synchronized at the same time.And now with the emergence of a large number of collection points,whether the use of data mining related technology to build the forecast model,the researchers are concerned about the problem.The purpose of this paper is to analyze the increasing environmental data,find the appropriate data mining technology,and then explore the better atmospheric environmental pollutant prediction model.The research process is based on six steps(data understanding,data preprocessing,traditional model,model evaluation,model explain,comparative optimization model).This study explores the new methods of forecasting,which is a useful complement to the traditional prediction model.Second section is about foreign case studies,data analysis and data mining techniques.US AirNow,providing the public with easy access to the country's overall air quality information and air quality index(AQI).The Australian Air Quality Forecasting System(AAQFS)is a front-end display frame that is used to predict the next day's air quality,which is input for weather and emission data,and is output for time-based prediction of air quality,currently used in Melbourne,Sydney and Adelaide.This paper discusses the technical methods of data mining.In this study,data preprocessing was used(data selection,integration,filtering,sampling,cleaning and conversion)technology to analyze the data,through multiple linear regression analysis(The analysis of the relationship between the attribute values in the same data,automatically produce models that can predict future attribute values)can be used to establish the prediction model.Third section explains data source,data structure and data preprocessing.Data is captured from public website,for this purpose,data crawler software has been used to process data.Data structure is consisting of data types and content.In data preprocess steps that consist on data cleaning and handling of missing data,statistical properties of the data are mean,standard deviation,variance,hypothesis tests etc.Analyzing data that has not been carefully screened for such problems can produce misleading results.After data preprocessing,the results will be applied to develop module.In the fourth section,traditional multiple linear regression models is studied,analysis the experiment of traditional model,optimization of multiple linear regression model,in this reason,optimization is divided into three parts such as correlation coefficient test,F test and t test.As far as linear regression is concerned,this module has been developed,and its result also compare with traditional module,then improve accuracy of module is achieved.Fitting degree of the optimized multiple linear models can reach 0.957 with the introduction of other pollutant concentrations and meteorological parameters.The fitting ratio of other pollutants and meteorological parameters was improved by 0.552 compared with the pure meteorological parameters.Improved multi-linear regression prediction model input parameters and input method changes have a greater impact on the program.The fifth part is the summary and the conclusion,in the past decade,atmospheric science has tried to use multiple linear regressions,regression tree and neural network and other statistical models to real-time prediction of air quality,but the amount of data was not enough so that the traditional linear model efficiency is not high.In this research improved the traditional multiple linear regression model and build a better prediction model.The data is analyzed by data preprocessing method,multiple linear regression analysis used to establish the prediction model,and then the correlation coefficient test,F test and t test are used to test the accuracy of the model.Finally,the forecast results are applied to the verification data.The experiment result show that the optimize multi linear regression model is better than the traditional linear regression model.The improved multiple linear regression models found that the PM2.5 concentration parameter had the greatest effect on the PM 10 concentration parameter.