Table 1.
Composition of the different subsamples used for the study and scores of the different classifiers.
| RotClass sample | PollFlag sample | PeriodSel sample | MissingParam sample | |
|---|---|---|---|---|
| Number of input stars | 21 707 | 315 Type 1 CP/CB candidates + 315 non-candidates | 14 562 | 1724 |
| RotClass accuracy (%) | 94.2/97.6 | – | – | 77.4 |
| PollFlag sensitivity (%) | – | 99.4 | – | 50 ( * ) |
| PeriodSel accuracy (%) | – | – | 95.3/99.5 | 79.4 |
| Global accuracy (%) | – | – | 92.1/96.9 | – |
Notes. The 315 non-candidate stars used in the PollFlag sample are randomly chosen in the PeriodSel sample at each training. For RotClass and PeriodSel, the metrics chosen to express the score is the accuracy while it is the sensitivity for PollFlag. When two scores are given, the first and the second one corresponds to the score before and after visual inspection, respectively. ( * )There is only 4 Type-1 CP/CB candidates in this sample, among which, only 2 have been correctly classified yielding 50% probability.
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.