Open AccessApplying Random Forest Classification to Ultracool Dwarf Discovery in Deep Surveys. II. Color Classification with PanSTARRS, 2MASS, UKIDSS, and WISE Photometry
Author(s) -
Eduardo Gauna Gutierrez,
Arantxa Mendiola Maytorena,
Zijie Gong,
Adriava Vega,
Carlos Verdaguer,
Christian Aganze,
Christopher Danner,
Adam J. Burgasser
Publication year - 2022
Publication title -
research notes of the aas
Language(s) - English
Resource type - Journals
ISSN - 2515-5172
DOI - 10.3847/2515-5172/ac6522
Subject(s) - random forest , photometry (optics) , classifier (uml) , pattern recognition (psychology) , artificial intelligence , physics , computer science , astrophysics , stars
We evaluate color-based classifiers in a synthesis of Pan-STARRS, 2MASS, UKIDSS, and AllWISE catalogs to identify ultracool dwarfs (UCDs). Using the Best et al. compilation of UCDs and a sample of background sources as our training set, we constructed a two-tier random forest model to segregate UCDs from non-UCDs and sort them into spectral subgroups. We also developed a regressor model to infer numerical classifications. Our classifier models achieved accuracies of 97%–99%, while our regressor model achieved a classification accuracy of 0.64 subtypes for classifications M5–T8. We applied these models to a 7 deg 2 region with overlapping survey data and identified 336 UCD candidates, of which 26 are previously identified UCDs and 17 are extragalactic sources.