Metallicity determination in galactic open star clusters by exploring multiwavelength surveys

Abstract

We developed models to estimate atmospheric parameters ( T _eff , log g , [Fe/H]) for 154 members of open clusters using the LightGBM machine learning algorithm, based on data from Gaia, the multi-band J-PLUS catalog, and CatWISE. Through feature engineering, we reduced the initial set of 153 features to 62, 65 and 76 for T _eff , log g , and [Fe/H], respectively, and the models achieved mean absolute errors of, respectively, 43 K, 0.055 dex, and 0.061 dex. Accounting for uncertainties in the input astrometric and photometric data, cluster metallicity errors were kept below 0.25 dex. This methodology is readily adaptable to other multi-band photometric surveys, such as S-PLUS and J-PAS, providing a robust framework for exploring the metallicity properties of additional clusters.