Table 1 Advantages and limitations of the existing works.
References | Methods | Advantages | Limitations |
|---|---|---|---|
Wu et al.15 | CNN with LSTM | Improved the system’s accuracy in forecasting traffic congestion | Harder to interpret and explain decisions |
Chen et al.16 | EMD with LSTM | Combining EMD for pre-processing and LSTM for temporal analysis leads to improved congestion prediction accuracy | Requires expertise in both time-series decomposition techniques |
Wang et al.17 | DG-CGRU | Focus on the most relevant spatial and temporal features | Require extensive hyperparameter tuning |
Bai et al.18 | CNN with LSTM | Scalable and versatile model | Increases the model’s complexity |
Majumdar et al.19 | Univariate and multi-variate lstms | Can handle complex, and large-scale traffic network | Needs large volumes of high-quality |
Alsubaiet al.20 | CNN with LSTM- IAO | By optimizing traffic flow and reducing congestion, sustainability in smart cities was promoted | Sensitivity to Parameter Initialization |
Chen et al.21 | DBN | Better at capturing complex patterns in large datasets, | Heavily reliant on the availability of high-quality |
Ata et al.22 | ABPN | Can adjust and refine their performance over time | May overfit to historical traffic data |
Zhai et al.23 | Bilstm | Handles non-linear relationships in traffic data | Difficult in areas with limited technological infrastructure |
Elfar et al.24 | Three ML models | Provided accurate short-term congestion predictions | High Computational Costs |
Mohanty et al.43 | Graph-CNN-LSTM | Can capture complex temporal and spatial relationships | Require significant computational resources |
