Photometric Classification of quasars from RCS-2 using Random Forest
Aims. Construction of a new quasar candidate catalog from the Red-Sequence Cluster Survey 2 (RCS-2), identified solely from photometric information using an automated algorithm suitable for large surveys. The algorithm performance is tested using a well-defined SDSS spectroscopic sample of quasars and stars. Methods. The Random Forest algorithm constructs the catalog from RCS-2 point sources using SDSS spectroscopically-confirmed stars and quasars. The algorithm identifies putative quasars from broadband magnitudes (g, r, i, z) and colours. Exploiting NUV GALEX measurements for a subset of the objects, we refine the classifier by adding new information. An additional subset of the data with WISE W1 and W2 bands is also studied. Results. Upon analyzing 542,897 RCS-2 point sources, the algorithm identified 21,501 quasar candidates, with a training-set-derived precision (the fraction of true positives within the group assigned quasar status) of 89.5% and recall (the fraction of true positives relative to all sources that actually are quasars) of 88.4%. These performance metrics improve for the GALEX subset; 6,530 quasar candidates are identified from 16,898 sources, with a precision and recall respectively of 97.0% and 97.5%. Algorithm performance is further improved when WISE data are included, with precision and recall increasing to 99.3% and 99.1% respectively for 21,834 quasar candidates from 242,902 sources. We compile our final catalog (38,257) by merging these samples and removing duplicates. An observational follow up of 17 bright (r < 19) candidates with long-slit spectroscopy at DuPont telescope (LCO) yields 14 confirmed quasars. Conclusions. The results signal encouraging progress in the classification of point sources with Random Forest algorithms to search for quasars within current and future large-area photometric surveys.