Multi-Station Air Quality Analysis and Prediction in Beijing Based on Deep Learning

With the increasingly severe problem of air pollution, it poses a serious challenge to human health, economic development, and environmental sustainability. How to effectively address this issue through accurate air quality analysis and prediction has become a focus of attention in recent years. This article is based on air quality data from 12 nationally controlled monitoring stations in Beijing from 2013 to 2017. In terms of space, the Air Quality Index (AQI) decreases with increasing distance from the city center, showing a trend of low in the north and high in the south. In terms of time, the peak of AQI occurs at 21:00 and the lowest value occurs at 7:00, which is closely related to people's daily activities. In summer, AQI is relatively good and more stable, while in winter, due to the use of heating equipment and fireworks during the Spring Festival, AQI is poor and fluctuates greatly. Through Pearson correlation analysis, PM2.5 and PM10 are the main pollutants that affect AQI, and the impact of meteorological conditions on AQI is relatively small. Although it cannot be immediately manifested, it cannot be ignored in the long-term trend. Finally, Utilizing the Long Short-Term Memory (LSTM) neural network model from deep learning to predict future AQI, with an Mean absolute error (MAE) of 0.7787 and a root mean square error (RMSE) of 0.9719 for 1 to 3 days, indicating high prediction accuracy. This study provides a scientific basis for government air quality management and citizens' daily lives.