EUVIMEDEUVIMED entdecken
Open Access · CC BY-SA 4.0

Evaluation of Tree-Based Ensemble Models Using Socioeconomic Data

Riechie, Ferry Vincenttius Ferdinand, Feliks Victor Parningotan Samosir

ComTech: Computer, Mathematics and Engineering Applications · 16.09.2026 · 17 (2) · 99-116

DOI: 10.21512/comtech.v17i2.14076 ↗

Kurz gefasst

Fachartikel in der Zeitschrift ComTech: Computer, Mathematics and Engineering Applications, online erschienen am 16.09.2026. Verfasst von Riechie und 2 weiteren Autor:innen. Die Arbeit nennt 34 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-SA 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Air quality significantly impacts public well-being and environmental sustainability, with socioeconomic factors known to influence air quality levels. However, existing AQI prediction studies predominantly rely on meteorological or pollutant variables, while the predictive capability of socioeconomic indicators remains underexplored. This study evaluates the effectiveness of tree-based ensemble learning techniques, including bagging, boosting, stacking, and multi-layer stacking, to predict air quality index (AQI) in Singapore using monthly socioeconomic indicators (34 variables) collected in Singapore from April 2014 until July 2024, encompassing transportation, tourism, industrialization, and price-related variables. Various tree-based algorithms, including Random Forest, ExtraTrees, XGBoost, and LightGBM, were evaluated. Stacking and multi-layer stacking models were also implemented with Linear Regression as the meta-learner. Feature engineering guided by the Pearson correlation coefficient was employed to optimize input selection. To optimize performance, hyperparameter tuning and model architectures were explored. The stacking model achieved the best predictive performance, with MAE, RMSE, and MAPE values of 13.68, 22.73, and 0.1359, respectively. Bagging models outperformed boosting models on the small dataset, while multi-layer stacking added computational complexity without significant performance gains. The results indicate that stacking captures complementary predictive patterns in socioeconomic data more effectively than deeper ensemble architectures, which offer limited benefits under data-constrained conditions. Furthermore, the findings show that socioeconomic variables provide sufficient predictive information to support AQI forecasting. Overall, this study highlights the potential of integrating socioeconomic data and offers methodological insights for future air quality research and applications.

Zitieren

APA

Riechie, Ferdinand, F. V., & Samosir, F. V. P. (2026). Evaluation of Tree-Based Ensemble Models Using Socioeconomic Data. ComTech: Computer, Mathematics and Engineering Applications, 17(2), 99-116. https://doi.org/10.21512/comtech.v17i2.14076

Vancouver

Riechie, Ferdinand FV, Samosir FVP. Evaluation of Tree-Based Ensemble Models Using Socioeconomic Data. ComTech: Computer, Mathematics and Engineering Applications. 2026;17(2):99-116. doi:10.21512/comtech.v17i2.14076

BibTeX
@article{Riechie2026,
  author = {Riechie and Ferdinand, Ferry Vincenttius and Samosir, Feliks Victor Parningotan},
  title = {Evaluation of Tree-Based Ensemble Models Using Socioeconomic Data},
  journal = {ComTech: Computer, Mathematics and Engineering Applications},
  year = {2026},
  volume = {17},
  number = {2},
  pages = {99--116},
  doi = {10.21512/comtech.v17i2.14076}
}
Kostenlos starten →