Answers are grounded in this paper and cite their sources.
Air pollution is recognized as a major public health risk worldwide, with fine particulate matter (PM2.5) identified as a key contributor to respiratory and cardiovascular diseases (World Health Organization, 2021). Accurate prediction of PM2.5 concentrations is essential for effective public health planning and environmental policymaking. Traditional statistical models have shown limitations in handling complex non-linear interactions between pollutants and meteorological variables. In this context, machine learning (ML) models, particularly ensemble methods like Random Forest, have gained popularity due to their predictive robustness and ability to capture intricate feature interactions (Park and Kim, 2025; Joharestani et al., 2019). This study focuses on predicting PM2.5 concentrations in Seoul using Random Forest Regressor (RFR), utilizing air quality data from the year 2021. Although a longer historical dataset was available, 2021 data was selected for its completeness and consistency across all key pollutant measurements. After data cleaning, feature engineering, and outlier removal, the model was trained and optimized via hyperparameter tuning. The optimized model achieved a Root Mean Squared Error (RMSE) of 4.88 and an R-squared value of 0.72. The results demonstrate the potential of Random Forest-based models for urban air quality forecasting. Future work will integrate meteorological parameters and explore advanced deep learning architectures to enhance predictive performance.
No takes yet. Share an insight, caveat, or question.
Abdallah et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: