ANALISIS PERBANDINGAN GRADIENT BOOSTING, LSTM, DAN AUTOFORMER DALAM FORECASTING KUALITAS UDARA SEBAGAI INDIKATOR DAMPAK PERUBAHAN IKLIM DI KOTA PALANGKARAYA
DOI:
https://doi.org/10.47111/jti.v20i2.27809Keywords:
air quality forecasting, LSTM, XGBoost, Autoformer, PM2.5, ISPUAbstract
Forest and peatland fires cause sharp fluctuations in Palangkaraya's air
quality. This study compares XGBoost, Long Short-Term Memory
(LSTM), and Autoformer for one-day-ahead forecasting of daily
PM2.5, CO, O3, and SO2 concentrations. A total of 950 daily
observations from September 2023 to May 2026 were obtained from
the Air Quality Open Data Platform and processed using backward
filling, logarithmic transformation, differencing, temporal feature
engineering, and a seven-day sliding window. Performance was
evaluated using MAE, RMSE, and R2. LSTM was the most consistent
model, attaining R2 values of 0.609 for PM2.5, 0.925 for O3, and 0.785
for SO2, while XGBoost performed best for CO with an R2 of 0.961.
Autoformer was less effective for most pollutants because of the limited
dataset and short sequence. LSTM was subsequently implemented in a
web system to produce H+1 forecasts and ISPU categories.
Downloads
References
[1] World Health Organization, WHO Global Air Quality Guidelines, Geneva: WHO, 2021.
[2] H. J. J. Hilly et al., ‘Review of scientific research on air quality and environmental health risk and impact for
PICTS,’ Science of the Total Environment, vol. 942, 2024, doi:10.1016/j.scitotenv.2024.173628.
[3] T. Chen and C. Guestrin, ‘XGBoost: A scalable tree boosting system,’ Proc. KDD, 2016,
doi:10.1145/2939672.2939785.
[4] A. Khumaidi, R. Raafi, and I. P. Solihin, ‘Pengujian algoritma Long Short Term Memory untuk prediksi
kualitas udara dan suhu Kota Bandung,’ Jurnal Telematika, vol. 15, no. 1, 2020.
[5] H. Wu, J. Xu, J. Wang, and M. Long, ‘Autoformer: Decomposition transformers with auto-correlation for
long-term series forecasting,’ NeurIPS, 2021/2022, arXiv:2106.13008.
[6] H. Hua et al., ‘The impact of data imputation on air quality prediction problem,’ PLoS ONE, vol. 19, no. 9,
2024, doi:10.1371/journal.pone.0306303.
[7] R. M. West, ‘Best practice in statistics: The use of log transformation,’ Annals of Clinical Biochemistry, vol.
59, no. 3, pp. 162-165, 2022.
[8] J. Leites, V. Cerqueira, and C. Soares, ‘Lag selection for univariate time series forecasting using deep learning:
An empirical study,’ arXiv:2405.11237, 2024.
[9] T. Chai and R. R. Draxler, ‘Root mean square error (RMSE) or mean absolute error (MAE)?’ Geoscientific
Model Development, vol. 7, pp. 1247-1250, 2014.
[10] D. Chicco, M. J. Warrens, and G. Jurman, ‘The coefficient of determination R-squared is more informative in regression analysis evaluation,’ PeerJ Computer Science, vol. 7, 2021, doi:10.7717/PEERJ-
CS.623.
[11] Kementerian Lingkungan Hidup dan Kehutanan RI, Peraturan Menteri LHK Nomor 14 Tahun 2020 tentang
Indeks Standar Pencemar Udara, 2020.
[12] T. S. Jaya, ‘Pengujian aplikasi dengan metode black-box testing boundary value analysis,’ Jurnal
Informatika: Jurnal Pengembangan IT, vol. 3, no. 2, 2018.
[13] M. R. Muhaimin, D. M. Karina, and A. B. Krisna, ‘Prediksi kualitas udara Malang menggunakan metode
Gradient Boosting Regression,’ Digital Transformation Technology, vol. 4, no. 2, pp. 937-942, 2024.
[14] B. Lim and S. Zohren, ‘Time-series forecasting with deep learning: A survey,’ Philosophical Transactions
A, 2021, doi:10.1098/rsta.2020.0209.
[15] M. Z. Wicaksono and L. E. Astrianty, ‘Sistem prediksi kualitas udara menggunakan algoritma LSTM,’ vol.
6, no. 7, pp. 831-840, 2025
