Air Quality Prediction Based On Long And Short Term Memory Neural Network Model

The degree of air quality is closely related to the concentration of pollutants in the air.The concentration of air pollutants in a given time and area can be influenced by many factors.The concentration of air pollutants in a given time and area can be influenced by many factors.The main reason is the amount of man-made pollutants discharged from fixed and mobile sources,such as vehicles,production emissions from industrial enterprises,residential heating and other factors.Meanwhile,geographical appearance and meteorological information of cities also have an important influence on air quality.The effective prediction of air quality changes in a specific geographical environment over a period of time not only helps the government to adjust environmental policies,but also provides a scientific basis for preventing air pollution.In order to evaluate air quality more effectively and increase its application value,the following work is studied in this paper:1)Correlation analysis of Air Quality Index(AQI)sequence and five factors,namely humidity,temperature,wind speed,visibility and pressure,was conducted to obtain the degree of correlation between AQI and meteorological factors.When the correlation coefficient is positive,it indicates that AQI is positively correlated with meteorological factors,and AQI increases with the increase of meteorological factors;When the correlation coefficient is negative,it indicates that AQI is negatively correlated with meteorological factors,and AQI decreases with the increase of meteorological factors.2)Air quality data of Beijing(PM2.5,PM10,SO2,NO2,O3 and CO)and meteorological data(humidity,temperature,wind speed,visibility and pressure)were used as input and AQI as output.On this basis,based on SVM were used respectively,BP neural network and LSTM of three kinds of neural network model for multiple input factor to predict the air quality,using the RMSE,MAE,MAPE and R~2four indexes of three kinds of prediction model were analyzed.Experimental comparison shows that the Long and Short Term Memory Neural Network can predict air quality more accurately.3)An air quality prediction system was designed using Python3.5 and Django frameworks.First of all,demand analysis is carried out for the function of air quality prediction system,the system is divided into five functional modules:user login,user management,data acquisition and cleaning,data query and statistics,and air quality prediction,the system with the SVM and BP neural network and LSTM three models based on neural network model,and by choosing LSTM neural network as the final prediction result,finally,the system implementation.