Applying Machine Learning to Determine Galaxy Membership in Galaxy Clusters
Presenter: Stephane Werner
Title: Applying Machine Learning to Determine Galaxy Membership in Galaxy Clusters
Date/Time: Thursday, June 20th, 12:10 PM
Abstract: Galaxy clusters are linked to many scientific questions related to galaxy evolution and cosmology, and are crucial for understanding these puzzles. Selecting galaxy memberships is a critical step when analyzing the properties of galaxy clusters, and incorrect selection can lead to misleading results. Recently, there have been significant advances in using machine learning algorithms to select cluster members. In Werner et al. (2022), we present a galaxy cluster catalog using data from the Southern Photometric Local Universe Survey (S-PLUS). We take advantage of its set of 12 filters in broad and narrow bands, which have led to improved photometric redshifts. We tested different machine learning algorithms to select member galaxies for our cluster catalog using a training sample built with spectroscopic redshifts and found that the Stochastic Gradient Boosting method yields better results in terms of purity and completeness. Member galaxies and interlopers have distinct properties statistically and we use these properties (such as galaxies’ colors, distance from the cluster center, and redshift offset) as features that are ranked by importance. This generates a completeness of 92.1±1.9% and purity of 85.7±2.3% for the training set. Finally, we applied this technique to 628 galaxy clusters at zphot≤0.23 and selected their member galaxies.