Traffic flow prediction can guide the rational layout of land use. Accurate traffic flow prediction can provide an important basis for urban expansion planning. This paper introduces a personalized lightweight federated learning framework (PLFL) for traffic flow prediction. This framework has been improved and enhanced to better accommodate traffic flow data. It is capable of collaboratively training a unified global traffic flow prediction model without compromising the privacy of individual datasets. Specifically, a spatiotemporal fusion graph convolutional network (MGTGCN) is established as the initial model for federated learning. Subsequently, a shared parameter mechanism of federated learning is employed for model training. Customized weights are allocated to each client model based on their data features to enhance personalization during this process. In order to improve the communication efficiency of federated learning, dynamic model pruning (DMP) is introduced on the client side to reduce the number of parameters that need to be communicated. Finally, the PLFL framework proposed in this paper is experimentally validated using LPR data from Changsha city. The results demonstrate that the framework can still achieve favorable prediction outcomes even when certain clients lack data. Moreover, the communication efficiency of federated learning under this framework has been enhanced while preserving the distinct characteristics of each client, without significant interference from other clients.

Accurate short‐term metro passenger flow prediction is critical for urban transit management, yet existing methods face two key challenges: (1) privacy risks from centralized data collection and (2) limited capability to model spatiotemporal dependencies. To address these issues, this study proposes a federated learning framework integrating convolutional neural networks (CNNs) and bidirectional gated recurrent units (BIGRU). Unlike conventional approaches that require raw data aggregation, our method facilitates collaborative model training across metro stations while keeping data stored locally. The CNN is employed to extract spatial patterns, such as passenger correlations between adjacent stations, while the BIGRU captures bidirectional temporal dynamics, including peak‐hour evolution. This architecture effectively eliminates the need for sensitive data sharing. We validate the framework using real‐world datasets from Shenzhen Metro, and our key innovations include a privacy‐preserving mechanism through federated parameter aggregation, joint spatial‐temporal feature learning without the need for raw data transmission, and enhanced generalization across heterogeneous stations.