Table 2
The table compiles the Mean Absolute Error (MAE) results achieved in this study for transit time forecasting and compares them with those obtained in earlier works that employed the DBM framework.
| Study | Model | Validation method | Test size | MAE [h] |
|---|---|---|---|---|
| Napoletano et al. (2018) | P-DBM | Hold-out | 14 | 9.1 |
| Hold-out | 100 | 16.8 | ||
| Dumbović et al. (2018) | DBEM | Hold-out | 25 | 14.3 |
| Paouris et al. (2021) | DBEM | Hold-out | 16 | 14.31 ± 2.18 |
| Napoletano et al. (2022) | P-DBM | Hold-out | 100 | 16.3 |
| This work (ensemble approach) | P-DBM | 4-fold CV | Slow – 17 [×4] | 10.3 ± 3.4 |
| Fast – 3 [×4] | 6.6 ± 0.7 | |||
| This work (individual approach) | P-DBM | 4-fold CV | M-I slow – 17 [×4] | 9.8 ± 4.1 |
| M-I fast – 3 [×4] | 7.9 ± 3.2 | |||
| MCMC slow – 18 [×4] | 11.1 ± 3.1 | |||
| MCMC Fast – 5 [×4] | 10.7 ± 7.7 | |||
| Liu et al. (2018) | Support vector machines | Best hold-out | 37 | 5.9 |
| Wang et al. (2019) | Convolutional neural network | 10-fold CV | 22 [×10] | 12.4 |
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.