ST-AGRNN: A Spatio-Temporal Attention-Gated Recurrent Neural Network for Traffic State Forecasting

Algorithm 1

Training of ST-AGRNN.

Input: The training epoch ; the historical traffic state ; the traffic graph ; the window size of historical traffic state ; the predicted length of traffic state ;

Output: Learned ST-AGRNN model

1: Initialization parameter ;

2: Data processing;

3: For do

4: Select real historical data ;

5: Select real future data ;

6: Input real historical data and the traffic graph into GCN and GRU to get ;

7: Input into attention to get ;

8: Use DeepWalk on and get the embedding result ;

9: Concatenate and _,;

10: Optimize by minimizing the loss function;

11: End for