Review Article
A Review on Deep Sequential Models for Forecasting Time Series Data
Table 6
A summary of published articles that used deep sequential models in traffic flow predictions.
| | Ref. | Method | Application | Results obtained | Metrics used | Data unit |
| | [83] | LSTM (TCC + LSM) | Traffic flow | Showed the efficacy and robustness of the proposed methods | MSE | Every 5 and 10 minutes | | [84] | Bidirectional LSTM, LSTM, and LSTM_BILSTM hybrid | Traffic (traffic flow prediction) | The suggested strategy outperformed both examined methods for accuracy and stability | RMSE = 16.7, MSE = 12.6, R2 = 0.86, MAPE = 6.3 | Every 5 minutes | | [85] | 1DCNN-LSTM | Traffic flow | The prediction impact of the proposed model demonstrated a faster convergence rate and increased forecast accuracy. | MAE, MSE, RMSE | Daily | | [86] | BRNN, LSTM, GRU | Traffic flow | The BRNN model forecasted velocity at three, six, and twelve time steps, BRNN the most advanced option, and provides with the lowest MAE and RMSE | MAE, RMSE | Every 20 minutes | | [87] | AHK-FOLSTM | Traffic speed prediction | FOLSTM determines the exact traffic speed from anomalous traffic data in order to overcome nonlinear features. | MSE, MAPE, MAE, and RMSE | 5 minutes |
|
|
