因果先验引导的交通流时空预测模型

侯越,王艳,董雨璇,赵建国,张琳栋,马静,宿增强

Causal prior-guided spatiotemporal traffic flow prediction model

Yue HOU,Yan WANG,Yuxuan DONG,Jianguo ZHAO,Lindong ZHANG,Jing MA,Zengqiang SU

表 1 不同模型在3个交通流数据集上的预测准确度

Tab.1 Prediction accuracy of different models on three traffic flow datasets

模型 PeMSD4 PeMSD8 成都-滴滴 MAE RMSE R2 MAE RMSE R2 MAE RMSE R2 TCN 23.06 36.03 0.9456 19.22 30.16 0.9546 2.79 4.05 0.8234 LSTM 22.14 36.11 0.9469 20.17 31.50 0.9537 2.78 4.04 0.8263 DCRNN 22.28 34.29 0.9166 16.74 25.63 0.9288 2.78 4.09 0.5818 ASTGCN 20.44 32.81 0.9193 16.49 25.36 0.9285 2.35 3.49 0.6920 SCINet 19.75 31.69 0.9598 16.50 25.71 0.9679 2.25 3.45 0.6905 STG-NCDE 20.29 31.36 0.9509 15.83 24.89 0.9708 2.32 3.44 0.6932 DeepSTUQ(23) 19.15 31.36 0.9339 15.04 23.97 0.9388 2.29 3.44 0.6933 BGCN(24) 18.83 31.73 0.9594 14.82 23.84 0.9732 2.35 3.34 0.7523 FGI(24) 19.31 31.68 0.9282 14.93 23.64 0.9396 2.19 3.31 0.7142 GINTRIP(24) 18.66 31.01 0.9572 14.91 24.05 0.9738 2.29 3.37 0.7075 CP-STN 18.36 30.23 0.9678 14.34 23.60 0.9699 2.14 3.18 0.8469