Traffic Flow Prediction based on Time Information

摘要

With the increasing number of urban vehicles, traffic congestion has become increasingly serious. Predicting traffic flow and take corresponding measures, which plays a pivotal role in easing traffic congestion. At present, many researchers have done a lot of research on the temporal and spatial correlation of traffic flow in the traffic network. Meanwhile, the traffic flow has a strong periodicity. If the periodicity is represented by time information, then the time information contains rich traffic flow information. Based on this, this paper introduces the time information into the ConvLSTM network and proposes BT-ConvLSTM (Based on Time ConvLSTM) network model, the traffic flow prediction accuracy is improved by using this model.