Graph Neural Network For Air Quality Prediction

Abstract

Air quality monitoring is important for public health and environmental ability. The study suggests a Attention Temporal Graph Convolutional Network (AT-GCN) for air quality prediction in Madrid, integrating Attention mechanisms, Gated Recurrent Units (GRUs), and Graph Convolutional Networks (GCNs). The AT-GCN model effectively treats asymmetrical data sources, including meteorological, traffic and pollution dataset, which increases the future of accuracy. It was trained on air quality data from January to June 2019 and tested on data from January to June 2022. The results were evaluated using RMSE, MAE, and correlation coefficient (R) metrics. Comparative analysis against baseline models such as LSTM, TGCN and horror highlighted GCNS better performance. The results show that AT-GCN effectively predict air quality, making it a valuable tool for environmental monitoring and policy design. By taking advantage of deep learning techniques, this model provides accurate prognosis, provides assistance to authorities and residents and makes informed to reduce pollution and improve air quality control.