Spectroscopic QUasar Extractor and redshift (z) Estimator squeze – I. Methodology

Abstract

ABSTRACT We present Spectroscopic QUasar Extractor and redshift (z) Estimator squeze, a software package to classify quasar spectra and estimate their redshifts. squeze is a random forest classifier operating on the parameters of candidate emission peaks identified in the spectra. We test the performance of the algorithm using visually inspected data from BOSS as a truth table. Only 4 per cent of the sample (∼6800 quasars and ∼11 520 contaminants) is needed for converged training in recommended choices of the confidence threshold (0.2 < pmin < 0.7). For an operational mode that balances purity and completeness (pmin = 0.32), we recover a purity of $97.40\pm 0.47{{\ \rm per\ cent}}$ ($99.59\pm 0.06{{\ \rm per\ cent}}$ for quasars with z ≥ 2.1) and a completeness of $97.46\pm 0.33{{\ \rm per\ cent}}$ ($98.81\pm 0.13{{\ \rm per\ cent}}$ for quasars with z ≥ 2.1). squeze can be used to obtain an ≈100 per cent pure sample of z ≥ 2.1 quasars (with ≈97 per cent completeness) by using a confidence threshold of pmin = 0.7. The estimated redshift error is $1500{\rm \, km\,s^{ -1}}$ and we recommend that squeze be used in conjunction with an additional step of redshift tuning to achieve maximum precision. We find that squeze achieves the necessary performance to replace visual inspection in BOSS-like spectroscopic surveys of quasars with subsequent publications in this series exploring expectations for future surveys and alternative methods.