A Method For Short-Term Traffic Flow Forecasting Based On GCN-LSTM

Spatial‐temporal correlation graph convolutional networks for traffic forecasting

“…To address this, we introduce the gated temporal convolution and perform dilated convolution with a proper dilation rate to enlarge the receptive field along the timeline (see Figure 4). Compared with previous work like RNN‐based approaches [46, 47], the gated convolution unit has a lighter structure and takes less time to calculate. Given the whole input data $$X\in \mathbb {R}^{T\times N\times d\times C}$$, the gated temporal convolution operation can be represented as follows: $$$$\begin{equation} Y_{output}=\psi {\left(\Phi _{1} \ast X+b_{1}\right)} \odot \sigma {\left(\Phi _{2} \ast X+b_{2}\right)}, \end{equation}$$$$where Φ 1 and Φ 2 are two 1D dilated convolution operations.…”

“…To address this, we introduce the gated temporal convolution and perform dilated convolution with a proper dilation rate to enlarge the receptive field along the timeline (see Figure 4). Compared with previous work like RNN-based approaches [46,47], the gated convolution unit has a lighter structure and takes less time to calculate. Given the whole input data X ∈ ℝ T ×N ×d ×C , the gated temporal convolution operation can be represented as follows:…”

Reducing hysteresis and over‐smoothing in traffic estimation: A multistream spatial‐temporal graph convolutional network

“…16 The GCN expands convolution operations to non-European space by using a graph structure, which can correspond to the characteristics of road networks. 17 Therefore, GCNs have recently become widely used in traffic estimation research. In addition, RNNs, long short-term memory (LSTM), and gated recurrent units (GRUs) can be used to learn temporal dependencies to achieve better estimation results.…”

“…Convolutional neural networks (CNNs) explore the spatial correlations of traffic road networks in European space 16 . The GCN expands convolution operations to non‐European space by using a graph structure, which can correspond to the characteristics of road networks 17 . Therefore, GCNs have recently become widely used in traffic estimation research.…”

Traffic Flow Prediction and Analysis in Smart Cities Based on the WND-LSTM Model

“…Based on the traffic flow model and the accurate space-time mode of each vehicle, the designed route guidance algorithm provides real-time and effective route planning services for end users, allows each vehicle to find the route that best meets its requirements at the lowest user cost, reduces the concentrated load on the road, and reduces the causes of traffic congestion to a certain extent. This reduces the causes of traffic congestion, reduces the pressure of traffic management departments, and saves the waste of human and material resources [ 11 ].…”