Accurate forecasting of air quality is essential for mitigating public health risks and supporting environmental policy decisions. This research proposes the development of a spatio-temporal prediction model that integrates both spatial correlations among monitoring stations and temporal patterns in pollutant variations.
By combining historical air quality measurements with meteorological, geographic, and emission source data, the model aims to capture complex dependencies across space and time. Advanced machine learning architectures such as graph neural networks and recurrent models are explored to improve prediction accuracy and robustness.
Expected Outcomes
The expected outcome is a predictive framework capable of providing fine-grained, short- and medium-term air quality forecasts, enabling timely interventions and more effective urban air quality management. Data sources include PurpleAir, OpenAQ, IQAir, BMKG (Indonesia national meteorological service), and AQICN.