Marín Díaz, Gabriel2026-07-082026-07-082026Marín Díaz, G. (2026). Fuzzy and Explainable AI for CMB Polarization Segmentation: Regional Stability Under Controlled Perturbations. Mathematics, 14(13), 2269. https://doi.org/10.3390/math141322692227-7390https://hdl.handle.net/11268/17258The cosmic microwave background (CMB) contains key information about the early Universe, particularly through its polarization structure. This work proposes a Fuzzy and Explainable Artificial Intelligence framework (FAS-XAI) for the regional analysis of CMB polarization using Planck SMICA data. From the Stokes components 𝑄 and 𝑈, the polarization amplitude 𝑃 and the scalar polarization modes 𝐸 and 𝐵 are derived. Regional features are then extracted over a HEALPix grid, considering only polarization-valid regions defined by the Planck polarization mask. Fuzzy C-Means identifies four interpretable polarization regimes: high-polarization structured regions, 𝐸-dominated medium-polarization regions, 𝐵-enhanced medium-polarization regions, and low-polarization regions. An XGBoost-SHAP layer is used to explain the resulting fuzzy memberships. XGBoost accurately reproduces the memberships, with 𝑅2>0.98 for all clusters, while SHAP confirms the physical relevance of amplitude-related features and the 𝑙𝑜𝑔(𝐵/𝐸) balance. Finally, controlled perturbations in 𝑃 and 𝑙𝑜𝑔(𝐵/𝐸) reveal a globally robust fuzzy structure with localized sensitivity. The proposed framework provides an interpretable methodology for studying regional CMB polarization patterns and their stability under controlled perturbations.engAttribution 4.0 Internationalhttp://creativecommons.org/licenses/by/4.0/Computación y tecnologíajournal article10.3390/math14132269open accessAstrofísicaInteligencia artificialAnálisis estadísticoGoal 4: Quality educationGoal 9: Build resilient infrastructure, promote sustainable industrialization and foster innovation