The 20th and 21st joint COTA International Conference of Transportation Professionals (CICTP 2020-21)

In the field of modern traffic flow forecasting, short-term traffic flow forecasting is a top priority because it has real-time prediction results, so it is directly applied to advanced traffic information systems and advanced traffic management systems to give walkers and managers. Provide dynamic, real-time and effective information. Based on the time series characteristics of traffic volume, this paper proposes a combined traffic flow prediction model of convolutional neural network(CNN) and gated recurrented unit (GRU) network which is based on deep learning. The CNN model is used to mine the parameters of traffic flow detection, and the time series features of traffic flow are mined by GRU model to realize short-term traffic prediction. The experimental results show that the combined model has an improved fit of 8.41% over the traditional long-short memory network (LSTM) model, which is an effective short-term traffic forecasting model.

CICTP