Pemanfaatan Pemodelan Machine Learning dalam Memprediksi Parameter Kualitas Udara Nitrogen Dioksida (NO2) Berdasarkan Algoritma Extra Trees Regression di DKI Jakarta

Abstract

The study examined the air quality in DKI Jakarta, mainly on the nitrogen dioxide (NO₂) parameters, using the Extra Trees Regression model to predict the NO₂ index. The study used the 2022 NO₂ time series data, which showed no significant long-term trends and indicated stationary and random data. Periodogram, histogram, and Q-Q plot analysis showed normal distribution with minor deviations. No significant autocorrelation was found between actual NO₂ data and model data, indicating the possibility of white noise. Evaluation of models with parameters such as MASE, MAE, RMSE, MAPE, SMAPE, and R₂ showed good model performance. An R² value of 73.14% indicates a model's ability to explain actual data variability. Although the Extra Trees Regression model follows seasonal patterns, there is an inconsistency between actual values and predictions at some points. This indicates a potential overfitting or difficulty in capturing specific data patterns. This research provides modeling information that is suitable for predicting air quality in DKI Jakarta.