Answers are grounded in this paper and cite their sources.
Recent observations can support short-term air quality prediction, but temporal persistence complicates the interpretation of high predictive agreement. We applied an attention-based long short-term memory (attention-LSTM) model to hourly observations from two urban monitoring stations in Heyuan during 2019–2023. SHapley Additive exPlanations (SHAP) were used to examine model-specific feature contributions. Reported test-set R² values ranged from 0.9796 to 0.9877 for PM2.5 and from 0.9921 to 0.9975 for O3. Corresponding RMSE values were 1.28–2.02 µg/m³ and 1.77–2.83 µg/m³. Predictions followed the displayed temporal patterns, while errors varied across concentration intervals. Recent pollutant observations received substantial SHAP importance, with humidity and temperature also contributing to O3 predictions. The analysis describes local prediction behavior and statistical feature associations. These associations do not establish pollutant formation mechanisms.
No takes yet. Share an insight, caveat, or question.
Wu et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: