Classification of Wolf Rayet stars using Ensemble-based Machine Learning algorithms

TL;DR

This paper presents an ensemble machine learning approach using XGB to classify and subtype Wolf-Rayet stars in the Milky Way based on infrared data, achieving high accuracy and discovering new candidates.

Contribution

The study introduces an XGB-based classifier for WR stars that outperforms other models and can identify new WR candidates and subtypes from large infrared datasets.

Findings

XGB classifier achieves 86% detection rate for WR stars.

The model accurately differentiates WR subtypes with over 60% accuracy.

Detected 58 new WR candidates in the Milky Way.

Abstract

We develop a robust Machine Learning classifier model utilizing the eXtreme-Gradient Boosting (XGB) algorithm for improved classification of Galactic Wolf-Rayet (WR) stars based on Infrared (IR) colors and positional attributes. For our study, we choose an extensive dataset of 6555 stellar objects (from 2MASS and AllWISE data releases) lying in the Milky Way (MW) with available photometric magnitudes of different types including WR stars. Our XGB classifier model can accurately (with an 86\% detection rate) identify a sufficient number of WR stars against a large sample of non-WR sources. The XGB model outperforms other ensemble classifier models such as the Random Forest. Also, using the XGB algorithm, we develop a WR sub-type classifier model that can differentiate the WR subtypes from the non-WR sources with a high model accuracy ($>60\%$). Further, we apply both XGB-based models to a selection of 6457 stellar objects with unknown object types, detecting 58 new WR star candidates and predicting sub-types for 10 of them. The identified WR sources are mainly located in the Local spiral arm of the MW and mostly lie in the solar neighborhood.