Table 4
Quality assessment of the XGBoost models generated during iterations 1 and 2.
Iteration 1 | Iteration 2 | |||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Assessment | DS0 | DS1 | DS2 | DS3 | DS4 | DS5 | DS6 | DS7 | DS0 | DS1 | DS2 | DS3 | DS4 | DS5 | DS6 | DS7 |
m etric | 170 | 155 | 146 | 146 | 146 | 28 | 28 | 28 | 170 | 155 | 146 | 146 | 146 | 28 | 28 | 28 |
µerr (prot) | 0.0535 | 0.0796 | 0.0898 | 0.0617 | 0.0425 | 0.0549 | 0.0452 | 0.0760 | 0.0423 | 0.0775 | 0.0849 | 0.0607 | 0.0424 | 0.0560 | 0.0458 | 0.0722 |
µerr (prot) | 0.0364 | 0.0549 | 0.0556 | 0.0475 | 0.0409 | 0.0480 | 0.0427 | 0.0532 | 0.0300 | 0.0549 | 0.0528 | 0.0463 | 0.0405 | 0.0489 | 0.0425 | 0.0499 |
acc10 | 0.923 | 0.859 | 0.854 | 0.889 | 0.896 | 0.884 | 0.892 | 0.861 | 0.931 | 0.861 | 0.868 | 0.897 | 0.896 | 0.884 | 0.895 | 0.873 |
RMSE (train/test) | 0.872 | 0.741 | 0.686 | 0.551 | 0.619 | 0.629 | 0.584 | 0.979 | 0.547 | 0.752 | 0.574 | 0.467 | 0.511 | 0.590 | 0.393 | 0.650 |
2.22 | 2.66 | 2.84 | 2.13 | 2.08 | 2.18 | 2.307 | 2.69 | 2.10 | 2.66 | 2.84 | 2.09 | 2.10 | 2.17 | 2.33 | 2.65 | |
MAE (train/test) | 0.377 | 0.452 | 0.425 | 0.339 | 0.384 | 0.378 | 0.357 | 0.575 | 0.241 | 0.453 | 0.351 | 0.289 | 0.321 | 0.345 | 0.251 | 0.395 |
0.803 | 1.27 | 1.34 | 1.03 | 1.07 | 1.08 | 1.13 | 1.28 | 0.719 | 1.29 | 1.28 | 0.995 | 1.07 | 1.08 | 1.12 | 1.22 | |
![]() |
0.963 | 0.947 | 0.942 | 0.955 | 0.948 | 0.952 | 0.938 | 0.943 | 0.967 | 0.947 | 0.942 | 0.957 | 0.948 | 0.953 | 0.937 | 0.945 |
Notes. All quality measures are presented with three significant figures. The header indicates the number of predictors in each data set. The first two rows show the mean absolute value of the relative residuals measured on the period and frequency regimes; the next row corresponds to the interval-based accuracy defined in Sect. 3.3, with a width of 10% of the reference values; the last three rows show the RMSE and the MAE days, both computed on the training and testing sets, and the adjusted coefficient of determination (calculated on the testing set). The best results for each metric (except for RMSE and MAE) are shown in bold.
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.