Enhancing Safety and Reducing Congestion in Smart Cities Using CrowdLOC-S with Graph Attention Networks

The growing complexity of metropolitan areas demands intelligent solutions capable of successfully tackling issues associated with safety and congestion. This study offers a Graph Attention Network (GAT) combined with CrowdLOC-S data to improve safety and reduce congestion in smart cities. The proposed approach utilizes graph-based learning to effectively capture spatial and temporal connections in crowd-sourced data, surpassing standard models in critical measures including accuracy, F1 score, and recall. The model reached an accuracy of 98.6%, an F1 score of 0.985, and a recall of 0.987, demonstrating substantial enhancements compared to prior models with accuracies of 92.5% and 89.3%. The results illustrate the effectiveness of the GAT model in enhancing urban management systems, establishing this method as a feasible alternative for real-time traffic and public safety monitoring. This research enhances smart city management and facilitates future progress in intelligent urban systems.