Traffic Flow Analysis and Prediction Based on Spatial-Temporal Data: A Case Study of North Cross Channel in Shanghai

Accurate traffic flow prediction is the key to achieving efficient traffic control and guidance. This paper focuses on traffic flow prediction considering temporal and spatial factors. To capture periodical dependence and spatial correlation, a Conv1D-LSTM prediction model was developed by combining one-dimensional convolution and pooling with the long short-term memory (LSTM) network. The developed model was validated with real-world traffic data sets of the Shanghai North Cross Channel. The analytical results of this case show that it is feasible to apply the developed Conv1D-LSTM model to predict traffic flow. Compared with support vector regression and seasonal autoregressive integrated moving average, this model has better prediction accuracy. This paper also employed the data at different road sections for testing and disclosed a better prediction performance. Ultimately, the Conv1D-LSTM-based traffic flow prediction model with superior accuracy and acceptable stability is expected to enhance the deployment of advanced transportation management systems.