IMADUDDIN, ADAM ROBBANI (2026) PERBANDINGAN PERFORMA LSTM, GRU DAN HYBRID DENGAN XGBOOST UNTUK PREDIKSI KUALITAS UDARA DI DKI JAKARTA. S1 thesis, Universitas Mercu Buana Jakarta.
|
Text (HAL COVER)
COVER.pdf Download (485kB) | Preview |
|
![[img]](https://repository.mercubuana.ac.id/style/images/fileicons/text.png) |
Text (BAB I)
bab 1.pdf Restricted to Registered users only Download (40kB) |
|
|
Text (BAB II)
bab 2.pdf Restricted to Registered users only Download (266kB) |
|
|
Text (BAB III)
bab 3.pdf Restricted to Registered users only Download (80kB) |
|
|
Text (BAB IV)
bab 4.pdf Restricted to Registered users only Download (400kB) |
|
|
Text (BAB V)
bab 5.pdf Restricted to Registered users only Download (93kB) |
|
|
Text (DAFTAR PUSTAKA)
Daftar Pustaka.pdf Restricted to Registered users only Download (97kB) |
|
|
Text (LAMPIRAN)
lampiran.pdf Restricted to Registered users only Download (600kB) |
Abstract
The increasing concentration of PM2.5 pollutants in DKI Jakarta has become a significant environmental issue that affects public health. Therefore, an accurate air quality prediction system is needed to support mitigation efforts and early warning mechanisms. This study aims to compare the performance of Long ShortTerm Memory (LSTM), Gated Recurrent Unit (GRU), and hybrid models integrated with Extreme Gradient Boosting (XGBoost) for predicting PM2.5 concentrations in DKI Jakarta. This research utilizes historical air quality and meteorological data obtained from the Copernicus Atmosphere Monitoring Service (CAMS) through the Open-Meteo service. The data undergoes several preprocessing stages, including feature engineering, normalization, and time series sequence formation using the sliding window approach. Subsequently, four prediction models are developed, consisting of LSTM, GRU, Hybrid LSTM-XGBoost, and Hybrid GRU-XGBoost. The performance of each model is evaluated using several metrics, including Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), and R². The results indicate that the GRU model achieves better performance than LSTM in capturing PM2.5 concentration patterns. Furthermore, the integration of XGBoost in the hybrid approach improves prediction performance by applying residual correction to the outputs of the base deep learning models. Among all tested models, the Hybrid LSTM-XGBoost model provides the best overall performance, demonstrating its potential as a foundation for developing an air quality prediction and early warning system. Keywords: PM2.5, Air Quality, Deep Learning, LSTM, GRU, XGBoost. Peningkatan konsentrasi polutan PM2.5 di wilayah DKI Jakarta menjadi salah satu permasalahan lingkungan yang berdampak terhadap kesehatan masyarakat. Oleh karena itu, diperlukan sistem prediksi kualitas udara yang mampu memberikan informasi secara akurat sebagai dasar mitigasi dan peringatan dini. Penelitian ini bertujuan untuk membandingkan performa model Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), serta model hybrid dengan Extreme Gradient Boosting (XGBoost) dalam memprediksi konsentrasi PM2.5 di DKI Jakarta. Penelitian ini menggunakan data historis kualitas udara dan faktor meteorologi yang diperoleh dari Copernicus Atmosphere Monitoring Service (CAMS) melalui layanan Open-Meteo. Data diproses melalui tahapan preprocessing, feature engineering, normalisasi, serta pembentukan data time series menggunakan metode sliding window. Selanjutnya, dilakukan pengembangan beberapa model prediksi, yaitu LSTM, GRU, Hybrid LSTMXGBoost, dan Hybrid GRU-XGBoost. Evaluasi performa model dilakukan menggunakan metrik Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), dan R². Hasil penelitian menunjukkan bahwa model GRU memiliki performa yang lebih baik dibandingkan LSTM dalam memprediksi pola perubahan konsentrasi PM2.5. Selain itu, penerapan pendekatan hybrid dengan XGBoost mampu meningkatkan akurasi prediksi melalui mekanisme koreksi residual terhadap hasil prediksi model dasar. Berdasarkan hasil perbandingan, model Hybrid LSTM-XGBoost menghasilkan performa terbaik dibandingkan model lainnya dan memiliki potensi untuk dikembangkan sebagai dasar sistem prediksi serta peringatan dini kualitas udara. Kata kunci: PM2.5, Kualitas Udara, Deep Learning, LSTM, GRU, XGBoost.
Actions (login required)
 |
View Item |
