Leveraging the Graph Convolutional Attention Hybrid Model for Real-Time Worker Safety Risk Prediction in the Gig Economy Risk Assessment and Mitigation Strategies

Worker safety is a critical concern in the dynamic and rapidly evolving gig economy. This study proposes a Graph Convolutional Attention Hybrid Model to predict real-time safety risks for gig workers by leveraging interconnected data on workers, tasks, and environmental factors. The dataset, collected from 1,247 gig workers, includes task characteristics, worker profiles, environmental conditions, and historical safety incidents. A graph-based approach utilized to model the relationships between nodes (workers, tasks, environmental factors) and edges, representing task dependencies or spatial proximity. The model achieved a remarkable accuracy of 98.81%, with a precision of 96.78%, recall of 97.34%, and AUC-ROC of 99.12%. These results demonstrate the hybrid model's ability to dynamically assign importance to critical risk factors using graph-based learning and attention mechanisms. The findings indicate that integrating graph convolutional networks (GCNs) and attention layers provides significant improvements over baseline models. The system also supports real-time inference, offering actionable insights for risk mitigation. This research contributes to advancing worker safety prediction frameworks and highlights the potential of hybrid machine learning models in dynamic and high-risk environments.