A Short-term Traffic Flow Prediction Model Based on AutoEncoder and GRU
Dejun Chen, Hao Wang, Ming Zhong
Abstract
To solve the problem of low prediction accuracy and poor robustness due to the short-term prediction only adopts the time series of current link traffic flow and fails to consider the spatial relationship in traffic flow data, this paper proposes a hybrid deep learning method considering the spatialtemporal correlation of traffic flow called AutoEncoder Gated Recurrent Unit (AE-GRU). The model fully considers the impact of upstream and downstream links traffic flow of the current link, can effectively use the topological spatial relationship of the road network, reduce the computational complexity, and can avoid the problem of unstable prediction results caused by the introduction of low-correlation road data. The results show that this method has higher prediction accuracy than the traditional prediction model and it is an effective traffic flow prediction model.