Speaker
Description
We report the comparison of two machine learning clustering algorithms, ML-MOC and HDBSCAN, for determining star cluster membership. ML-MOC employs a combination of the k-Nearest Neighbors (k-NN) algorithm to find candidate members, followed by a Gaussian Mixture Model (GMM) for precise membership probability assignment. This makes it well-suited for regular, centrally concentrated clusters. In contrast, HDBSCAN is a density-based cluster-finding algorithm with a hierarchical structure, making it ideal for identifying extended features of arbitrary shapes. Using proper motion and parallax measurements from Gaia Data Release 3 (DR3), we applied these algorithms to two morphologically distinct targets: M67, a compact, well-defined open cluster, and M39, a cluster exhibiting extended tidal structures. Here we quantified the pros and cons of how each algorithm performed to uncover different clustering structures, e.g., in terms of completeness and reliability against field contamination, and how a combination of the two improves the membership determination.
| Participate the oral/poster presentation award competition | Yes |
|---|