In this page, you will find a description of the method for the extraction of photometric QSOs candidates described in the paper "Quasar candidates selection in the Virtual Observatory era" from D'Abrusco et al. submitted to MNRAS (preprint).
The inspiring principle of this work is the application of statistical and data-mining techniques to obtain a clustering of astronomical sources inside a photometric parameter space and fully characterize the distribution of different types of sources inside this parameter space. This concept has been applied to the problem of the selection of QSOs candidates from broadband photometric data by exploiting the availability of large spectroscopic bases of knowledge (BoK: i.e., samples of sources with a reliable classification).
The procedure for the extraction of candidates can be summarized as follows:
A BoK consisting of a sample of stellar sources with spectroscopic classification is clustered inside the colour parameter space. This BoK is drawn from the catalogue of photometric sources from where, at the end of the process, the new QSOs candidates will be extracted.
Several possible partitions of the distribution of sources of the BoK inside the colour space are produced by a combination of two clustering algorithm: PPS and NEC.
The members of each cluster of each different partition are labelled using the BoK classification.
Amongst all the possible partitions in the colour space, the one allowing the best separation between clusters populated mainly by confirmed QSOs ("successful" clusters) and clusters populated mainly by contaminants is considered.
The new candidates QSOs are selected as the photometric sources which are associated, in the colour space, to the "successful" clusters by a suitable distance definition.
The details of the method and algorithms can be found in the paper.
The catalogues of QSOs candidates extracted from the SDSS DR7 photometric survey can be downloaded here.