Alfasanah Zulfani, Niam M Zaim Husnun, Wardiani Sri, Ahsan Muhammad, Lee Muhammad Hisyam
Department of Statistics, Institut Teknologi Sepuluh Nopember, Indonesia.
Department of Mathematical Sciences, Universiti Teknologi Malaysia, Johor Bahru, Malaysia.
MethodsX. 2024 Dec 12;14:103107. doi: 10.1016/j.mex.2024.103107. eCollection 2025 Jun.
PM2.5 air pollution poses significant health risks, particularly in urban areas such as Jakarta, where concentrations frequently surpass acceptable levels due to rapid urbanization. This study addresses autocorrelation in air quality data and evaluates the monitoring performance of XGBoost and Support Vector Regression (SVR) models using Individual and Exponentially Weighted Moving Average (EWMA) Charts. PM2.5 levels were obtained from Jakarta's Air Quality Index. The findings reveal that the SVR model effectively manages autocorrelation, while the combination of XGBoost and the EWMA chart yielded superior monitoring performance. Specifically, this approach detected only one out-of-control (OOC) point in Phase II and none in Phase I, with identified shifts ranging from moderate to large. Overall, the XGBoost and EWMA chart integration offers a robust solution for precise air quality monitoring and minimizes false alarms. The identification of OOC points provides actionable insights by highlighting significant deviations in air quality data that may require immediate intervention. Key points:•SVR and XGBoost model regression was introduced to enhance forecasting accuracy.•EWMA chart based on XGBoost residuals has better monitoring results.
细颗粒物（PM2.5）空气污染带来重大健康风险，在雅加达等城市地区尤为如此，由于快速城市化，那里的PM2.5浓度经常超过可接受水平。本研究探讨空气质量数据中的自相关性，并使用个体和指数加权移动平均（EWMA）控制图评估极端梯度提升（XGBoost）和支持向量回归（SVR）模型的监测性能。PM2.5水平取自雅加达空气质量指数。研究结果表明，SVR模型能有效处理自相关性，而XGBoost与EWMA控制图相结合产生了更优的监测性能。具体而言，这种方法在第二阶段仅检测到一个失控（OOC）点，在第一阶段未检测到任何失控点，识别出的偏移从中度到大幅度不等。总体而言，XGBoost与EWMA控制图相结合为精确空气质量监测提供了强大的解决方案，并最大限度减少误报。识别OOC点通过突出空气质量数据中可能需要立即干预的显著偏差提供了可采取行动的见解。要点：•引入SVR和XGBoost模型回归以提高预测准确性。•基于XGBoost残差的EWMA控制图具有更好的监测结果。
