Fecha
2024-09-01
Editor/a
Director/a
Tutor/a
Coordinador/a
Prologuista
Revisor/a
Ilustrador/a
Derechos de acceso
info:eu-repo/semantics/openAccess
Título de la revista
ISSN de la revista
Título del volumen
Editorial
Citas
0 citas en 
0 citas en 
Resumen
Context. Investigating the formation of open clusters is essential for understanding the evolution of the MilkyWay’s stellar content. Various machine learning algorithms have been implemented to identify members within these clusters, however, the performance of these methods has not been properly analyzed. Aims. The main objective is to analyze and evaluate various algorithms and metrics used for membership identification in open clusters of stellar systems. This includes a detailed assessment of the performance, time and objectives of the algorithms in two distinct experiments.
Methods. The analysis of key algorithms such as HDBSCAN, StarGO, pyUPMASK and Miec using two different synthetic clusters that mimic the characteristics of Gaia catalog. The experiments differ in the number of objects, as the algorithms’ performance can be highly dependent on this factor.
Results. For numerous clusters, the best algorithm depended on the objectives of the classification: HDBSCAN for fast computing, pyUPMASK for complete selection of the members and Miec for a more balanced approach. For smaller clusters, Miec was the best algorithm in all metrics. Similarly, there is no best metric for all cases. Similarly, no single metric is ideal for all situations. For cases requiring a clean selection of members, contamination rate and PPV are recommended. For a more complete list of cluster members, TPR is useful, while MCC and LSR offer a more balanced approach.
Conclusions. The performance of the algorithm and the metrics selected depend on the objectives of the classification itself. Therefore, there is no universally best algorithm or metric, so it’s important to evaluate the goals and select the most suitable accordingly.
Descripción
Categorías UNESCO
Palabras clave
methods, algorithm and metric analysis, open clusters and associations: general, astrometry, photometry
Citación
Nieto Petinal, Isabel. Trabajo Fin de Máster: Comparative Analysis of Machine Learning Algorithms for Member Classification in Stellar Systems. Universidad Nacional de Educación a Distancia (UNED), 2024
Centro
E.T.S. de Ingeniería Informática
Departamento
Inteligencia Artificial