Development Of Air Quality Prediction And Visualization System Based On Deep Learning

The quality of air affects people's sense of well-being.In recent years,due to the rapid development of society,air pollution has become increasingly serious,which makes people in the quality of life and health and other aspects of a great threat.Among them,PM2.5particulate matter is more prominent.Too high PM2.5 concentration will lead to reduced visibility in cities,forming heavy haze weather,and seriously affecting people's transportation and social production.In addition,high PM2.5 concentration can cause respiratory diseases,cardiovascular diseases,and premature death of patients with heart and lung diseases.Therefore,establishing a scientific and accurate prediction model is a necessary prerequisite for controlling and preventing the harm caused by air pollution.Based on this,an IPSO-Bi LSTM neural network prediction model was proposed in this paper.Air quality factors and meteorological factors were selected as experimental data,and feature factors with high correlation with PM2.5 were screened out as feature input through grey correlation analysis.Bi LSTM was used as the basic model.The learning rate,Batchsize,epoch and node number of hidden layer of Bi LSTM neural network were optimized by PSO.Since PSO has some problems,such as low convergence accuracy and easy to fall into local extremum,discrete factor and time factor are introduced on the original basis,and the parameters of PSO are nonlinear changed.However,PSO algorithm still has its own defects,in order to further reduce the probability of falling into local extreme value,improve the performance of the algorithm.In this paper,WPA algorithm with strong global optimization ability is introduced.After improving step size parameters in WPA optimization process,Bi LSTM neural network model based on IWPA-PSO hybrid algorithm is proposed.By comparing with other models and conducting extended experiments,it is found that,IWPAPSO-Bi LSTM model has the best prediction performance,the highest prediction accuracy and stronger robustness.In order to improve the use value of data,the neural network model is integrated into the air quality visualization system based on JQuery,Spring Boot and other front-end frameworks,realizing the functions of user login and registration,historical data query,air quality prediction,data management and user management.It provides a new way for system users to understand air quality data,meteorological data information and popularize air pollution.