Name:	Masci, Frank
Email:	fmasci@ipac.caltech.edu
Institution:	Caltech
Title:	Machined-learned classification of eclipsing and pulsating variable stars detected by WISE
Topic:	Time Domain
Abstract:	We are currently contructing a catalog of variable and transient sources using candidates selected from the AllWISE Multiepoch Photometry Database. Here, we describe our classification methodology and performance for specific classes of periodic variable stars commonly detected from the first year of WISE observations. We compiled a truth set of 8273 variable stars from the literature and derived several metrics ("features") from their WISE light curves. The class-to-feature mappings were then used to train a machine-learned classifier to support the future classification of unknown variables. The classifier is based on the commercially-popular Random Forest method. For the three most common types of periodic variables identified by WISE: Algols, RR-Lyrae, and W Ursae Majoris, we obtain classification efficiencies of > 80% and purity levels of >~ 90%. This rivals the performance achieved in previous automated classification studies of periodic variable stars.