多分辨率趋势周期解耦交互的交通流预测

侯越,王甜甜,张鑫,尹杰

Traffic flow forecasting with multi-resolution trend period decoupling interaction

Yue HOU,Tiantian WANG,Xin ZHANG,Jie YIN

表 2 不同模型在2个数据集不同时间步长下的交通流预测准确度

Tab.2 Traffic flow prediction accuracy of different models at time steps in two datasets

模型 t/min PeMSD4 PeMSD8 MAE RMSE MAPE MAE RMSE MAPE HA 5 25 45 24.020 28.760 35.180 39.930 45.260 53.870 0.201 0.219 0.253 20.820 26.020 31.360 34.670 39.890 46.570 0.171 0.187 0.212 SVR 5 25 45 26.007 27.460 29.580 38.630 40.360 42.820 0.195 0.205 0.223 18.047 19.220 21.100 26.504 27.980 30.140 0.130 0.138 0.152 LSTM 5 25 45 22.721 24.082 24.401 37.487 39.394 39.883 0.159 0.168 0.170 20.008 22.244 22.322 31.837 35.799 35.724 0.125 0.134 0.139 CNN_LSTM 5 25 45 22.079 22.481 22.485 35.723 36.601 36.636 0.145 0.156 0.158 20.205 21.014 21.360 31.540 32.719 33.356 0.127 0.131 0.133 TCN 5 25 45 18.345 22.504 26.269 29.095 35.371 40.333 0.122 0.146 0.172 13.721 19.461 20.821 21.295 31.443 32.082 0.086 0.117 0.128 ASTGCN 5 25 45 17.737 20.927 23.342 28.343 33.045 36.429 0.108 0.125 0.138 14.241 18.000 20.531 21.787 27.842 31.581 0.090 0.105 0.116 SCINet 5 25 45 19.109 19.532 20.368 29.473 31.306 32.259 0.120 0.125 0.126 13.989 16.729 17.655 21.798 26.490 26.876 0.087 0.099 0.114 MTPDI 5 25 45 11.153 13.635 14.453 17.029 21.791 22.248 0.091 0.106 0.111 9.091 10.355 10.763 13.485 16.060 16.270 0.061 0.068 0.073