Air Quality Prediction Based on PITS

This study employs the Patch Independence for Time Series (PITS) model for air quality prediction using data from three Chinese monitoring stations collected from April 2019 to July 2021. The research compares PITS with traditional LSTM and Autoformer models in predicting six major pollutants (SO2, $\mathrm{NO}_{2}, \mathrm{PM}_{10}, \mathrm{PM}_{2.5}, \mathrm{O}_{3}, \mathrm{CO}$). PITS demonstrated superior performance across all metrics, achieving the lowest MAE (3.5392), RMSE (5.3346), and highest $\mathbf{R}^{\mathbf{2}}$ (0.8603). Through comprehensive analysis of predicted concentrations, O3 was identified as the primary pollutant in 28 out of 30 station-days. Further investigation using XGBoost and SHAP analysis revealed temperature and humidity as key influencing factors for O3 concentrations. These findings provide valuable insights for air quality management and pollution control strategies, particularly in addressing O3 pollution through temperature and humidity regulation.