Boost recall in quasi-stellar object selection from highly imbalanced photometric datasets - The reverse selection method

Open Access

Table 4

Evolution of dataset sizes and class imbalance at the beginning of each step of the reverse selection method.

	Step1	Step 2	Step 3	Step 4
N^train	22 525 474	138 505	18 568	5749
N^test	5 631 369	32 141	5237	1891
N^unclassified	2 639 184	1 324 967	107 193	81 853

	0.007%	1.102%	8.218%	26.544%
	0.007%	1.126%	6.912%	17.610%

Notes. The first three rows report the size of the training, test, and unclassified datasets. The last two rows report the fraction of high-z QSOs in the training and test datasets (this quantities are not available for the unclassified dataset).

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.