FARROS, Fauzia Dhiyaa' and Gernowo, Rahmat and Surarso, Bayu (2026) PEMODELAN PREDIKSI SPASIAL-TEMPORAL AIR QUALITY INDEX (AQI) DKI JAKARTA BERBASIS HYBRID CATEGORICAL BOOSTING REGRESSOR DAN ATTENTION MECHANISM. Masters thesis, UNIVERSITAS DIPONEGORO.

	Text 1. cover awal.pdf Download (250kB)
	Text 2. cover lengkap.pdf Restricted to Repository staff only Download (1MB)
	Text 3. BAB I.pdf Download (291kB)
	Text 4. BAB II.pdf Download (942kB)
	Text 5. BAB III.pdf Restricted to Repository staff only Download (3MB)
	Text 6. BAB IV.pdf Restricted to Repository staff only Download (2MB)
	Text 7. BAB V.pdf Restricted to Repository staff only Download (263kB)
	Text 8. Daftar Pustaka.pdf Download (278kB)
	Text 9. Lampiran.pdf Restricted to Repository staff only Download (3MB)

Abstract

Penelitian ini bertujuan mengembangkan model prediksi Indeks Kualitas Udara (AQI) di DKI Jakarta dengan mengkombinasikan algoritma CatBoost Regressor dan mekanisme Feature Attention Weighting yang memberikan bobot perhatian 0.5 untuk kondisi normal dan 1.0 untuk kondisi yang paling ekstrem. Dataset terdiri atas data ISPU DKI Jakarta dan data meteorologi GHCN-H NOAA periode 2020– 2024 dengan total 105.466 entri mencakup PM2.5, PM10, SO₂, CO, O₃, NO₂, suhu udara, suhu titik embun, kelembapan relatif, kecepatan angin, arah angin, curah hujan 24 jam, tekanan udara permukaan laut, dan jarak pandang. Model dilatih menggunakan sekuens waktu (1, 8, 12, 24, dan 48 jam) dan skema pembagian data 70:15:15 (train:validation:test), kemudian dibandingkan dengan tujuh model lain yaitu Linear Regression, Lasso Regression, LightGBM, FNN, CNN-LSTM, GRU- LSTM, dan Transformer. Hasil terbaik diperoleh pada jendela 48 jam dengan RMSE = 0.002, uji Friedman menunjukkan signifikansi statistik (p = 0.010) dan Pairwise Test mengkonfirmasi signifikansi statistik pada RMSE (χ² = 11.200; p = 0.024), MAE (χ² = 13.280; p = 0.010), dan MAPE (χ² = 13.280; p = 0.010). Analisis
spasial-temporal menunjukkan lonjakan PM₂.₅ dan PM₁₀ pada pagi dan sore akibat lalu lintas dan aktivitas industri, dengan konsentrasi tertinggi di daerah Lubang Buaya (PM₂.₅ = 287 µg/m³, PM₁₀ = 187 µg/m³) serta lonjakan meteorologi menunjukkan perubahan harian yang dipengaruhi kondisi cuaca. Fluktuasi ini memengaruhi konsentrasi variabel di Halim Perdanakusuma pada kelembapan, suhu, arah angin. Secara umum, CatBoost menunjukkan efektivitas, ketepatan, dan signifikansi statistik, menjadikannya pilihan tepat untuk pemodelan prediksi indeks kualitas udara DKI Jakarta.
Kata kunci : Air Quality Index (AQI), Categorical Boosting (CatBoost) Regressor, Attention Mechanism, Spasio-Temporal, PM2.5

This study aims to develop an Air Quality Index (AQI) prediction model for DKI Jakarta by integrating the CatBoost Regressor algorithm with a Feature Attention Weighting mechanism that assigns an attention weight of 0.5 for normal conditions and 1.0 for the most extreme conditions. The dataset comprises ISPU air-quality records from DKI Jakarta and GHCN-H NOAA meteorological data from 2020– 2024, totaling 105.466 entries and including PM2.5, PM10, SO₂, CO, O₃, NO₂, temperature, dew point temperature, relative humidity, wind speed, wind direction, precipitation 24 hour, sea level pressure, visibility. The model was trained using multiple time-sequence windows (1, 8, 12, 24, and 48 hours) and a 70:15:15 train– validation–test split, and subsequently benchmarked against seven comparative models: Linear Regression, Lasso Regression, LightGBM, FNN, CNN-LSTM, GRU-LSTM, and Transformer. The best performance was achieved at the 48-hour window with an RMSE of 0.002. The Friedman test indicated statistical significance (p = 0.010), and Pairwise tests further confirmed significant differences across RMSE (χ² = 11.200; p = 0.024), MAE (χ² = 13.280; p = 0.010), and MAPE (χ² = 13.280; p = 0.010). Spatiotemporal analysis revealed notable PM₂.₅ and PM₁₀ spikes during morning and evening hours driven by traffic and industrial activity, with the highest concentrations observed in Lubang Buaya (PM₂.₅ = 287 µg/m³; PM₁₀ = 187 µg/m³). Meteorological fluctuations also showed daily variability influenced by weather conditions, affecting humidity, temperature, and wind direction in the Halim Perdanakusuma area. Overall, CatBoost demonstrated strong effectiveness, high accuracy, and statistically significant performance, underscoring its suitability for AQI prediction modeling in DKI Jakarta.
Keywords : Air Quality Index (AQI), Categorical Boosting (CatBoost) Regressor, Attention Mechanism, Spasio-Temporal, PM2.5

Actions (login required)

View Item