Traffic Flow Prediction in Urban Built-Up Areas Using Deep Learning

Urban traffic flow prediction in built-up areas holds significant importance for traffic management and public safety. However, it is susceptible to spatial dependencies, temporal dependencies, and additional factors such as weather and unexpected events. This paper proposes a novel deep learning model, the Spatiotemporal Fusion ResNet (STF-ResNet), for traffic flow forecasting. By integrating four structurally identical residual networks with a fully connected network, the model addresses the limitation of single neural network architectures in simultaneously capturing spatiotemporal correlations. Comparative experiments were conducted using real-world taxi mobility data from New York City, evaluating the STF-ResNet against LSTM, CNN, and ST-ResNet models. The experimental results demonstrate that our model achieves higher predictive accuracy than other baseline models.