Monitoring Kualitas Udara PM₂,₅ Menggunakan Adaptive Exponentially Weighted Moving Average (AEWMA) Berbasis Residual XGBoost

Rahim, Yolanda (2025) Monitoring Kualitas Udara PM₂,₅ Menggunakan Adaptive Exponentially Weighted Moving Average (AEWMA) Berbasis Residual XGBoost. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6003232010-Master-Thesis.pdf - Accepted Version Restricted to Repository staff only Download (7MB) | Request a copy

Abstract

Kualitas udara merupakan faktor penting yang memengaruhi kesehatan manusia dan lingkungan. Pemantauan kualitas udara yang efektif diperlukan untuk mendeteksi perubahan konsentrasi polutan secara dini dan mendukung pengambilan keputusan yang tepat. Penelitian ini mengusulkan pendekatan pemantauan kualitas udara menggunakan diagram kontrol Adaptive Exponentially Weighted Moving Average (AEWMA) berbasis residual dari model prediksi Extreme Gradient Boosting (XGBoost). AEWMA dipilih karena sensitivitasnya terhadap perubahan kecil maupun besar, sementara XGBoost mampu menangani data deret waktu nonlinier dan mereduksi autokorelasi pada residual. Data PM₂,₅ harian dikumpulkan dari platform AQI untuk wilayah Jakarta pada periode Januari 2023 hingga Mei 2025. Pemodelan dilakukan dalam dua pendekatan: lag signifikan dan sliding window. Hasil evaluasi menunjukkan bahwa pendekatan lag signifikan menghasilkan residual yang lebih stabil dengan nilai MAPE lebih rendah, sedangkan pendekatan sliding window lebih adaptif terhadap dinamika data terbaru namun menghasilkan lebih banyak sinyal out of control (OOC). Integrasi XGBoost dan AEWMA terbukti efektif sebagai sistem pemantauan dini kualitas udara berbasis data. Hal ini didukung juga dengan hasil perbandingan terhadap diagram kontrol lainnya, di mana AEWMA khususnya pendekatan lag signifikan menunjukkan performa paling seimbang dalam mendeteksi anomali, dibandingkan EWMA (λ = 0.1) yang cenderung kurang adaptif, serta grafik individu yang memiliki batas kontrol terlalu lebar dan kurang peka terhadap perubahan kecil.
====================================================================================================================================
Air quality is a critical factor affecting both human health and the environment. Effective air quality monitoring is essential to detect early changes in pollutant concentrations and to support informed decision-making. This study proposes an air quality monitoring approach using an Adaptive Exponentially Weighted Moving Average (AEWMA) control chart based on the residuals of an Extreme Gradient Boosting (XGBoost) prediction model. AEWMA is chosen for its sensitivity to both small and large shifts, while XGBoost is capable of handling nonlinear time series data and reducing autocorrelation in residuals. Daily PM₂.₅ data were collected from the AQICN platform for the Jakarta area during the period from January 2023 to May 2025. Modeling was conducted using two approaches: significant lags and a sliding window. Evaluation results show that the significant lag approach produced more stable residuals with lower MAPE values, while the sliding window approach was more adaptive to recent data dynamics but generated more out-of-control (OOC) signals. The integration of XGBoost and AEWMA has proven effective as a data-driven early warning system for air quality monitoring.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Adaptive Exponentially Weighted Moving Average (AEWMA), Extreme Gradient Boosting (XGBoost), Pemantauan Kualitas Udara, Deret Waktu, PM₂,₅, Air Quality Monitoring, Time Series
Subjects:	H Social Sciences > HA Statistics > HA29 Theory and method of social science statistics H Social Sciences > HA Statistics > HA31.38 Data envelopment analysis. H Social Sciences > HD Industries. Land use. Labor > HD9980.5 Service industries--Quality control. Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines.
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49101-(S2) Master Thesis
Depositing User:	Yolanda Rahim
Date Deposited:	05 Aug 2025 12:19
Last Modified:	05 Aug 2025 12:19
URI:	http://repository.its.ac.id/id/eprint/127602

Actions (login required)

View Item