STGNN-TCN: Hybrid Model for Spatiotemporal Air Quality Prediction based on Spatio-Temporal Graph Neural Networks and Temporal Convolutional Networks

The development of correct air quality forecasting remains essential for the task of environmental monitoring as well as public health management practices. Traditional deep learning techniques find it difficult to analyse complex patterns of spatial along with temporal relationships in air quality datasets. The authors present a new hybrid deep learning structure which combines Spatio-Temporal Graph Neural Networks (ST-GNN) with Temporal Convolutional Networks (TCN) to address this issue. ST-GNN extracts valuable spatial features by building complex spatial relationship models which TCN uses to detect long-range temporal patterns in order to produce reliable time-series forecasts. A hyperparameter search using Bayesian Optimization helps to optimize model performance by speeding up convergence while reducing potential overfitting issues. The proposed framework shows superior performance through air quality data evaluation which demonstrates both enhanced predictive accuracy and improved computational efficiency when compared to traditional deep learning approaches. The putative evidence demonstrates that hybrid framework achieves strong scalability with reliability which makes it suitable as a tool for real-time air quality prediction systems. The research establishes a contemporary data-directed system for environmental monitoring which supports better decision outcomes in public health as well as pollution control activities.