Photometric classification of quasars from RCS-2 using Random Forest

Abstract

We describe the construction of a quasar catalog containing 91,842 candidates derived from analysis of imaging data with a Random Forest algorithm. Using spectroscopically-confirmed stars and quasars from the SDSS as a training set, we blindly search the RCS-2 (∼ 750 deg2) imaging survey. From a source catalogue of 1,863,970 RCS-2 point sources, our algorithm identifies putative quasars from broadband magnitudes (g, r, i, and z) and colours. Exploiting NUV GALEX mea- surements available for a subset 16,898 of these objects, we refine the classifier by adding NUV-optical colours to the algorithm’s search. An additional subset (comprising 13% of the source catalog) features WISE coverage; we explore the effect of including the W1 and W2 bands on the performance of the algorithm. Upon analysing all RCS-2 point sources, the algorithm identified 85,085 quasar candidates, with a training-set-derived precision (the fraction of true positives within the group assigned quasar status) of 90.4% and a recall (the fraction of true positives relative to all sources that actually are quasars) of 87.3%. These performance metrics improve for the subset with GALEX data; 6,556 quasar can- didates are identified with a precision and recall respectively of 96.9% and 97.3%. Algorithm performance is improved further still with the analysis of WISE data, with precision and recall further increasing to 99.3% and 99.2% respectively for 21,713 quasar candidates. Upon merging these samples and removing duplicates, we arrive our final catalog of 91,842 quasar candidates. An observational follow up of 17 bright (r < 19) potential quasars with long-slit spectroscopy at DuPont telescope (LCO) yields 13 confirmed quasars. Whilst this preliminary sample is small, it signals encouraging progress in the use of Random Forest algorithms to classify point sources for quasar searches within large-area photometric surveys such as the LSST.