From: Traffic flow prediction based on depthwise separable convolution fusion network
Method(L + W + E + T) | MAE | RAE (× 100) | RMSE | RRSE (× 100) | MAPE | R2 (× 100) | Trainable params |
---|---|---|---|---|---|---|---|
FFN(SConv1D + earlyTGTR) | 148.7 | 63.0 | 227.7 | 67.8 | 15.5 | 54.1 | 198101 |
FFN(SConv1D + middleTGTR) | 150.0 | 63.5 | 228.6 | 68.0 | 15.7 | 53.7 | 197641 |
FFN(SConv1D + lateTGTR) | 152.1 | 64.4 | 231.0 | 68.7 | 16.0 | 52.8 | 197641 |
FFN(SConv1D + none) | 151.6 | 64.2 | 229.3 | 68.2 | 15.9 | 53.4 | 196911 |
FFN(Conv1D + none) | 154.5 | 65.4 | 230.2 | 68.5 | 16.4 | 53.1 | 205821 |
DL-LSTM [19] | 160.8 | – | 233.5 | – | 16.7 | 51.7 | – |
DL-FC [19] | 152.6 | 64.6 | 232.1 | 69.1 | 16.1 | 52.3 | 238313 |