Table 2 Different ANN networks architectures for traffic volume prediction
From: ANN based short-term traffic flow forecasting in undivided two lane highway
Model | Algorithm | Hidden layer | hidden neurons | Transfer Function | Epochs | Learning | Step size/Mo | Training | Cross validation | Testing | |||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Min MSE (*104) | Final MSE (*104) | Min MSE (*104) | Final MSE (*104) | MSE | NMSE (*104) | MAE | R (%) | ||||||||
M1 | MLP | 1 | 4 | Tanh | 100 | LM | – | 5.9 | 5.9 | 12.5 | 10.37 | 4.66 | 47.99 | 1.93 | 98.8 |
M2 | MLP | 1 | 4 | Tanh | 200 | LM | – | 5.3 | 5.3 | 15 | 7.13 | 6.00 | 61.81 | 2.14 | 98.4 |
M3 | MLP | 1 | 4 | Tanh | 300 | LM | – | 4 | 4 | 21.1 | 20.94 | 3.25 | 33.49 | 1.26 | 98.7 |
M4 | MLP | 1 | 5 | Tanh | 100 | LM | – | 3.6 | 3.8 | 20.24 | 8.371 | 4.79 | 49.34 | 1.84 | 99.2 |
M5 | MLP | 1 | 5 | Tanh | 200 | LM | – | 2.8 | 2.8 | 16.05 | 16.866 | 19.8 | 204.4 | 3.73 | 93.7 |
M6 | MLP | 1 | 5 | Tanh | 100 | MO | 1/0.7 | 29.9 | 29.9 | 23.19 | 2.506 | 4.12 | 42.43 | 1.64 | 98.3 |
M7 | MLP | 1 | 5 | Tanh | 200 | MO | 1/0.7 | 27.3 | 27.3 | 25.8 | 3.3 | 7.10 | 73.12 | 2.06 | 96.6 |
M8 | MLP | 1 | 3 | SigmoidAxon | 100 | LM | – | 2.6 | 2.6 | 3.09 | 0.832 | 7.980 | 82.1 | 2.32 | 97.9 |
M9 | MLP | 1 | 3 | SigmoidAxon | 200 | LM | – | 1.6 | 1.8 | 2.63 | 0.263 | 17.66 | 181.8 | 3.13 | 92.7 |
M10 | MLP | 1 | 5 | SigmoidAxon | 100 | LM | – | 0.81 | 0.8141 | 3.73 | 1.883 | 15.012 | 154.55 | 3.44 | 95.6 |
M11 | MLP | 1 | 5 | SigmoidAxon | 200 | LM | – | 0.03 | 0.0358 | 4.92 | 13.479 | 151.9 | 1564.4 | 11.18 | 25.5 |
M12 | MLP | 1 | 7 | SigmoidAxon | 150 | LM | – | 0.36 | 0.366 | 4.7 | 2.89 | 0.816 | 8.4 | 0.77 | 99.8 |