J-PLUS DR3: Galaxy-Star-Quasar classification
First author: R. von Marttens
The Javalambre Photometric Local Universe Survey (J-PLUS) is a 12-band photometric survey using the 83-cm JAST telescope. Data Release 3 includes 47.4 million sources (29.8 million with $r \le 21$) on 3192 deg$^2$ (2881 deg$^2$ after masking). J-PLUS DR3 only provides star-galaxy classification so that quasars are not identified from the other sources. Given the size of the dataset, machine learning methods could provide a valid alternative classification and a solution to the classification of quasars. Our objective is to classify J-PLUS DR3 sources into galaxies, stars and quasars, outperforming the available classifiers in each class. We use an automated machine learning tool called {\tt TPOT} to find an optimized pipeline to perform the classification. The supervised machine learning algorithms are trained on the crossmatch with SDSS DR12, LAMOST DR7 and \textit{Gaia} DR3. We checked that the training set of about 570 thousand galaxies, one million stars and 220 thousand quasars is both representative and pure to a good degree. We considered 37 features: besides the twelve photometric bands with their errors, six colors, four morphological parameters, galactic extinction with its error and the PSF relative to the corresponding pointing. After exploring numerous pipeline possibilities through the TPOT genetic algorithm, we found that XGBoost provides the best performance: the AUC for galaxies, stars and quasars is above 0.99 and the average precision is above 0.99 for galaxies and stars and 0.94 for quasars. XGBoost outperforms the star-galaxy classifiers already provided in J-PLUS DR3 and also efficiently classifies quasars. We also found that photometry was very important in the classification of quasars, showing the relevance of narrow-band photometry.