.png){kind=icon}
.png){kind=icon}
Return
IoT-enabled air quality prediction using ensemble machine learning: a case study in San Salvador, El Salvador
.png){kind=icon}
PRE
AI
J
D
G
B
M
Y
O
C
W
DOI：10.1007/s11869-026-02031-3.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Air pollution poses a risk to human health in densely populated urban areas. In El Salvador, limited air quality monitoring infrastructure hinders the implementation of risk prevention policies regarding exposure to polluted air. This study presents the first work on predicting the Air Quality Index (AQI-UBA) using artificial intelligence, developed using data from an IoT network deployed in the Historic Center of San Salvador. The IoT network consists of three monitoring stations that collected parameters such as temperature, humidity, CO2, total volatile organic compounds (TVOC), and the AQI-UBA index over a period of 15.85 weeks. Nine machine learning algorithms are evaluated individually; these models are: Random Forest, CatBoost, GradientBoosting, XGBoost, LightGBM, AdaBoost, ExtraTrees, SVR, and Ridge Regression. Additionally, an ensemble model called SECGR is proposed, which integrates SVR, ExtraTrees, CatBoost, and GradientBoosting as base models; the predictions generated by these base models are combined using Ridge Regression as a meta-model to generate a final prediction. The results show that the SECGR model achieves an R2 value of 0.9970, an MAE of 0.0016, an RMSE of 0.0395, and a MAPE of 0.06%. These results enable predictions that can support real-time alerts, public health decision-making, and sustainable urban planning, strengthening air quality management in urban environments.
Keywords:
Air pollution
Machine learning
Ensemble learning
Environmental monitoring
IoT
Journal
A
IF:
0
Papers:
143
Citations:
0
Organization
Cited Papers