A Parallax Distribution-Based Gaussian Mixture Model for Membership Determination in Open Clusters of Solar Neighborhood
Rafli Rizaldi (a*) and Muhamad Irfan Hakim (a,b)

a) Astronomy Study Program, Faculty of Mathematics and Natural Sciences, Institut Teknologi Bandung
^raflirizaldigamer[at]gmail.com
b) Bosscha Observatory, Institut Teknologi Bandung

Abstract

This study presents a machine learning approach for determining membership of newly identified open clusters in the solar neighborhood using Gaia DR3 data. Previous works have employed k-nearest neighbors (kNN) and Gaussian Mixture Models (GMM) to improve cluster membership identification. However, applying this method to poorly studied clusters often leads to difficulties in defining reliable parameter ranges using kNN alone. To address this limitation, we introduce a three-component parallax distribution-based GMM during the initial stage of membership selection. In the initial stage, the kNN algorithm is used to determine the range of astrometric parameters from cluster data downloaded within a small search radius. This step is followed by a three-component GMM applied to the parallax distribution to improve the initial membership selection. Subsequently, a two-component GMM is applied to the one-dimensional distribution of Mahalanobis distance. This process is performed on data obtained with a larger search radius, using the parameter range derived from the previous step to probabilistically identify cluster membership. The method is applied to eight clusters: four well-studied control clusters (NGC 2360, Alessi 1, NGC 2099, and NGC 752) and four poorly studied clusters (OCSN 13, 22, 37, and 56). The proposed method effectively distinguishes cluster members from field stars and yields consistent results with existing literature. Furthermore, the parallax distribution-based GMM reveals nine new cluster candidates, two of which correspond to OCSN 57 and OCSN 61. This method shows promising potential for broader application in the discovery of new clusters using current and future Gaia data releases.

Keywords: Open cluster- Machine learning- Clustering- kNN- GMM

Topic: Galaxies and Cosmology