Prediction of the Air Quality Index in DKI Jakarta Province Using the CatBoost Method

Abstract

Air pollution in major cities like Jakarta continues to worsen due to various contributing factors, including unregulated industrial emissions, open waste burning, and the increasing number of private vehicles. This study aims to classify air quality levels based on the Air Pollution Standard Index (ISPU) using the CatBoost Classifier algorithm. The dataset comprises ISPU data from 2021 to 2024 sourced from Jakarta's public data portal, including parameters such as PM10, PM2.5, SO2, CO, O3, and NO2. After preprocessing and feature selection, the model was trained and evaluated using standard classification metrics. The CatBoost Classifier achieved high performance in major categories like “BAIK”, “SEDANG”, and “TIDAK SEHAT” with F1-scores exceeding 0.94. However, the “SANGAT TIDAK SEHAT” category could not be predicted accurately due to class imbalance. To address this, a hybrid model incorporating rule-based logic was employed, enabling accurate classification in the case of extreme pollution. The model also offers station-level predictions, supporting spatial analysis and early warning systems. The results demonstrate that the proposed approach provides a robust framework for air quality classification and real-time environmental monitoring.