Scientific Development Research

With the continuous expansion of urban areas and the increasing complexity of emergency management needs, multi-source heterogeneous data fusion has become an urgent issue in urban emergency management. This paper proposes a federated learning-based multi-source heterogeneous data fusion framework, aiming to integrate valuable information from diverse data sources including sensors, monitoring systems, and social media through distributed learning methods, thereby achieving precise and intelligent emergency response and decision-making. The framework fully considers data privacy protection by employing federated learning technology, enabling all participants to jointly train models without exchanging raw data, thus enhancing both data security and analytical effectiveness. Experimental results demonstrate that this framework can significantly improve the response efficiency and accuracy of emergency management systems, adapting to rapidly changing urban environments.

Urban emergency management; Multi-source data; Heterogeneous data; Federated learning; Data fusion