Kualitas udara di Jakarta semakin mengkhawatirkan akibat tingginya emisi kendaraan bermotor, aktivitas industri, dan faktor meteorologi yang memperburuk akumulasi polutan. Salah satu polutan yang menjadi perhatian utama adalah PM 2.5 (particulate matter ≤ 2,5 µm).Menurut WHO Global Air Quality Guidelines (2021) PM 2.5 ditetapkan sebagai indikator utama kualitas udara global karena dampaknya yang signifikan terhadap kesehatan. PM 2.5 mampu menembus hingga ke alveolus paru-paru dan masuk ke aliran darah, sehingga meningkatkan risiko penyakit pernapasan, kardiovaskular, stroke, hingga kematian dini. Penelitian ini berfokus pada pemodelan dan peramalan konsentrasi PM 2.5 di Jakarta dengan membandingkan dua metode, yaitu Autoregressive Integrated Moving Average (ARIMA) yang unggul dalam memodelkan pola linier deret waktu, dan Support Vector Regression (SVR) yang mampu menangkap hubungan non-linear melalui kernel trick. Data PM 2.5 harian diperoleh dari stasiun pemantauan resmi dan diolah melalui tahap pembersihan, normalisasi, serta pembagian data latih dan uji. Kinerja model dievaluasi menggunakan Root Mean Squared Error (RMSE) dan Mean Absolute Percentage Error (MAPE). Hasil penelitian menunjukkan bahwa SVR memberikan tingkat akurasi yang lebih tinggi dibandingkan ARIMA dalam memprediksi PM 2.5 di Jakarta, sehingga berpotensi digunakan sebagai basis sistem peringatan dini dan pengambilan kebijakan pengendalian polusi udara yang lebih efektif.

Air quality in Jakarta has become increasingly concerning due to high emissions from motor vehicles, industrial activities, and meteorological factors that exacerbate pollutant accumulation. One pollutant of primary concern is PM 2.5 (particulate matter ≤ 2.5 µm).According to the WHO Global Air Quality Guidelines (2021), PM 2.5 is designated as a key global air quality indicator due to its significant health impacts. PM 2.5 can penetrate deep into the alveoli of the lungs and enter the bloodstream, thereby increasing the risk of respiratory diseases, cardiovascular disorders, stroke, and premature death. This study focuses on modeling and forecasting PM 2.5 concentrations in Jakarta by comparing two methods: Autoregressive Integrated Moving Average (ARIMA), which excels in modeling linear time series patterns, and Support Vector Regression (SVR), which can capture non-linear relationships through the kernel trick. Daily PM 2.5 data were obtained from official monitoring stations and processed through cleaning, normalization, and splitting into training and testing sets. Model performance was evaluated using Root Mean Squared Error (RMSE) and Mean Absolute Percentage Error (MAPE). The results show that SVR achieved higher accuracy than ARIMA in predicting PM 2.5 in Jakarta, indicating its potential as the basis for early warning systems and more effective air pollution control policies.