A deep learning urban traffic congestion forecast model blending the temporal continuity and periodicity

Abstract

Traffic congestion has become an inevitable and difficult disease in the process of urban development, and it has also brought harm and hidden dangers to citizens' travel and urban development. The emergence of GCN solves the problem of capturing the spatial characteristics of urban road traffic. Based on this, we propose a new method that considers the periodicity of traffic patterns and builds a neural network model with multiple time scales to capture more detailed features. And the experiment proves that our model is better in predicting traffic congestion.