<![CDATA[Air pollution presents a substantial threat to human health, especially in urban areas like Jakarta, Indonesia, which ranked eleventh worldwide for poor air quality and urban pollution in mid-2025. This study is conducted with the objective of forecasting air quality over a designated future period by employing two advanced machine learning techniques: the Extreme Learning Machine (ELM) and its kernel-based variant, the Kernel Extreme Learning Machine (K-ELM). These methodologies are applied to predict the concentrations of five features of pollutants—PM10 (Particulate Matter), SO2 (Sulfur Dioxide), CO (Carbon Monoxide), O3 (Ozone), and NO2 (Nitrogen Dioxide)—which are critical indicators of environmental air quality and have significant implications for human health and environmental sustainability. Both methods are evaluated for their efficiency in time series regression, with a focus on training speed and generalization performance. The results demonstrate that the K-ELM model, especially when utilizing a Laplacian kernel, outperforms the standard ELM in predicting air quality based on the air quality index (AQI) dataset. Performance metrics indicate that K-ELM achieves superior accuracy, with an RMSE of 0.041, MSE of 0.002, MAE of 0.019, and an R-squared value of 0.898, confirming its effectiveness for air quality prediction in Jakarta. Furthermore, the Nemenyi post-hoc analysis across all metrics showed that K-ELM with the Laplacian kernel consistently achieved the highest rank and exhibited statistically significant improvements in multiple pairwise comparisons.]]>
