PM_2.5 Concentration Prediction And Web System Design Based On Decision Tree

At present,the situation of air pollution in China is becoming more and more serious,and the sharp decline of air quality has caused serious harm to people's health,and it also hinders the sustainable development of society and economy.It has been widely concerned by people in China and around the world to predict the concentration of PM_2.5 so as to monitor the air pollution and prevent serious pollution.Therefore,it is an important issue to propose effective models for accurate prediction of PM_2.5 concentration.In this paper,the PLS-M5P?Partial Least Square-M5P?model is proposed for PM_2.5 concentration prediction.The method first uses partial least squares?Partial Least Square,PLS?to analyze the air quality data,and obtains the main factors affecting the PM_2.5 concentration.Then,these air mass factors are used as the input vectors of the M5P algorithm.The model tree prediction model of PM_2.5concentration is obtained.Using the historical data of the air quality in the Haidian New Area of Beijing,the concentration level of PM_2.5 in the air is predicted.At the same time,the system is designed on the online network for the model discussed.The system is based on the J2EE,uses the Struts2 and other open source frameworks,and uses the HTML5 technology for the front display,The use of HTML5 enhances system interactivity,and also improves the user's network experience.The main research work included in this paper is as follows:1.extracting air quality data about PM_2.5 concentration:This paper takes the Haidian New Area in Beijing as the research area,which includes the time series data of daily air pollutants:PM_2.5?ug/m³?,PM₁₀?ug/m³?,sulfur dioxide?ug/m³?,nitrogen dioxide?ug/m³?,oxygen carbon?ug/m³?,ozone?ug/m³?,and temperature?°C?,relative humidity?%?,air pressure?mbar?,wind speed?m/s?,wind direction?Dir?.The weather data for 46 consecutive days from December 1,2014 to January 15,2015are selected as experimental data.The partial least squares?PLS?method is used to analyze the air quality data,and the main factors that affect the PM_2.5 concentration are PM₁₀,NO₂,CO,SO₂,humidity and wind speed as the important input parameters of the PM_2.5 prediction model.At the same time,the principal component analysis method is also used as a contrast.From the main factors extracted by partial least-squares method,the prediction accuracy of PM_2.5 concentration has been improved.2.Establishment of PM_2.5 Concentration Prediction Model Based on PLS-M5P:In order to solve the problem of low prediction accuracy and opaque prediction process in PM_2.5,the PLS-M5P model is proposed in this paper.Using the historical data of the air quality in the northern New Area of Beijing,Prediction of the future concentration level of PM_2.5 in the air.Partial least squares method was used to extract the main factors affecting PM_2.5 concentration.The main factors extracted were used as input vectors of M5P model tree and a prediction model was established.The experimental results show that,compared with the BP neural network and the PCA-M5P?Principal Component Analysis-M5P?model,the PLS-M5P model is more accurate for the prediction of PM_2.5 concentration level.In terms of air quality prediction,compared with the traditional prediction model,such as the BP neural model,the PLS-M5P model tree has the following advantages:?1?provide intuitive mathematical equations and provide more in-depth Understand the forecast result.?2?The tree graph generated using the PLS-M5P model shows the importance of the factors,and the establishment of the tree chart enables the decision-makers to understand the prediction process more clearly.?3?Modeling and prediction take a short time and always converge.?4?The prediction accuracy is higher.3.PM_2.5 online network prediction system design:In order to achieve real-time PM_2.5 forecasting and air pollution data display,established a PM_2.5 online network forecasting system.The system first uses Sql Server 2012 to build a database of air quality data,stores air quality data,and then builds Tomcat on a Web server for running and parsing front pages.Finally,use MyEclipse Java EE to develop the background and front page of the system.Insert the trained PLS-M5P decision tree model at the background.The HTML5 front page makes data requests.The Servlet processes the front data request.The background database extracts the required data and transfers it to the Tomcat server.Then Servlet retrieves the data.Data sent to the front page for display,real-time display of air quality data and PM_2.5 predicted value,and corresponding physical health and travel protection tips based on the air quality data and the predicted value of PM_2.5,It helps people get more convenient air pollution and life tips.