FAS-XAI for Interpretable Stellar Population Discovery: Fuzzy Memberships, Explainable Artificial Intelligence, and Spatial Analysis of Hipparcos Data
AI · 14.11.2026 · 7 (11) · 488
DOI: 10.3390/ai7110488 ↗Kurz gefasst
Fachartikel in der Zeitschrift AI, online erschienen am 14.11.2026. Verfasst von Gabriel Marín Díaz. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Stellar populations exhibit continuous physical transitions that are difficult to represent using discrete classification schemes. This work presents FAS-XAI (Fuzzy Analysis and Explainable Artificial Intelligence), a framework that combines fuzzy clustering, supervised reconstruction, explainable artificial intelligence, and interactive exploration. Using photometric and astrometric measurements from 49,273 Hipparcos stars, Fuzzy C-Means identifies stellar populations while preserving graded memberships and overlap between neighbouring groups. XGBoost provides an internal consistency check by reconstructing the dominant FCM assignments with 98.83% accuracy, while SHAP explains the supervised surrogate model. The framework identifies four stellar populations: hot luminous main-sequence stars, intermediate main-sequence stars, cool low-luminosity dwarfs, and red giant and evolved stars. The B-V colour index, absolute visual magnitude, and distance are the most influential variables, while proper motion has a smaller contribution. The association with classical spectral classifications provides an additional astrophysical consistency check for the identified fuzzy stellar groups. An interactive environment also allows population-level filtering and individual-star inspection. FAS-XAI provides a framework for discovering and interpreting continuous stellar population structures in large astronomical catalogues.
Zitieren
APA
Díaz, G. M. (2026). FAS-XAI for Interpretable Stellar Population Discovery: Fuzzy Memberships, Explainable Artificial Intelligence, and Spatial Analysis of Hipparcos Data. AI, 7(11), 488. https://doi.org/10.3390/ai7110488
Vancouver
Díaz GM. FAS-XAI for Interpretable Stellar Population Discovery: Fuzzy Memberships, Explainable Artificial Intelligence, and Spatial Analysis of Hipparcos Data. AI. 2026;7(11):488. doi:10.3390/ai7110488
BibTeX
@article{Daz2026,
author = {Díaz, Gabriel Marín},
title = {FAS-XAI for Interpretable Stellar Population Discovery: Fuzzy Memberships, Explainable Artificial Intelligence, and Spatial Analysis of Hipparcos Data},
journal = {AI},
year = {2026},
volume = {7},
number = {11},
pages = {488},
doi = {10.3390/ai7110488}
}
EUVIMED