A Census of Object Types and Redshift Estimates in the SDSS Photometric Catalog from a Trained Decision Tree Classifier

Abstract

We have applied ClassX, an oblique decision tree classifier optimized for astronomical analysis, to the homogeneous multicolor imaging data base of the Sloan Digital Sky Survey (SDSS), training the software on subsets of SDSS objects whose nature is precisely known via spectroscopy. We find that the software, using photometric data only, correctly classifies a very large fraction of the objects with existing SDSS spectra, both stellar and extragalactic. ClassX also accurately predicts the redshifts of both normal and active galaxies in SDSS. To illustrate ClassX applications in SDSS research, we (a) derive the object content of the SDSS DR2 photometric catalog and (b) provide a sample catalog of resolved SDSS objects that contains a large number of candidate AGN galaxies, 27,000, along with 63,000 candidate normal galaxies at magnitudes substantially fainter than typical magnitudes of SDSS spectroscopic objects. The surface density of AGN selected by ClassX to i~19 is in agreement with that quoted by SDSS. When ClassX is applied to the photometric data fainter than the SDSS spectroscopic limit, the inferred surface density of AGN rises sharply, as expected. The ability of the classifier to accurately constrain the redshifts of huge numbers (ultimately ~ 10^7) of active galaxies in the photometric data base promises new insights into fundamental issues of AGN research, such as the evolution of the AGN luminosity function with cosmic time, the starburst--AGN connection, and AGN--galactic morphology relationships.