Graph Neural Network-Based Optimization of IoT-Connected Smart Waste Systems in Sustainable Cities

Urban waste management has major difficulties because of rising trash volumes and ineffective collection systems, resulting in environmental and operational strains. This study presents a Graph Neural Network (GNN)-based optimisation framework for IoT-enabled smart waste systems to tackle these challenges. The system forecasts fill levels and optimises collection routes by modelling waste bins and their spatial interactions as graph nodes and edges, resulting in improved accuracy and efficiency. In comparison to conventional machine learning benchmarks such as Random Forest, SVM, and LSTM, the GNN model attained enhanced metrics, including 94.7% accuracy, 0.93 F1-score, and minimised error rates (MAE: 0.081, RMSE: 0.098). These improvements facilitate real-time, data-informed decision-making that reduces collection frequency, fuel usage, and carbon emissions. This study strengthens sustainable urban garbage management through the combination of graph-based learning with IoT, showcasing scalability and practical viability for smart city implementations.