Random Forests as a Viable Method to Select and Discover High-redshift Quasars

We present a method of selecting quasars up to redshift ?6 with random forests, supervised machine-learning method, applied Pan-STARRS1 and WISE data. find that, thanks the increasing set known quasars, we can assemble training that enables algorithms become competitive alternative other methods this redshift. candidate for range 4.8–6.3, which includes region around z = 5.5 where is difficult due their photometric similarity red brown dwarfs. demonstrate under our survey restrictions, reach high completeness (66% ± 7% below 5.6/ above 5.6) while maintaining selection efficiency (/). Our estimated via novel based on different distributions contaminants sky. The final catalog 515 candidates 225 quasars. predict contain additional new 5.6 above, make publicly available. Spectroscopic follow-up observations 37 led us discover 20 (18 at 4.6 ? 5.5, 2 ? 5.7). These are consistent predictions efficiency. argue forests lead higher because contains number objects would be rejected by common color cuts, including one newly discovered 5.7