<p>High-dimensional fuzzy data pose serious challenges to conventional Fuzzy Multi-Attribute Decision-Making (FMADM) methods, as the curse of dimensionality significantly degrades decision efficiency. In this context, integrating attribute reduction algorithms with decision-making methods is regarded as a key approach to improving performance. However, existing attribute reduction techniques often fail to adequately capture the complex nonlinear relationships among attributes and suffer from low reduction efficiency, limited accuracy, and insufficient robustness. To address these issues, this study proposes a novel Gaussian Kernel Function-based Unsupervised Attribute Reduction (GKFUAR) algorithm. The algorithm innovatively integrates an Attribute Self-Expression Function (ASEF) with a Triple Regularization Mechanism (TRM), which effectively overcomes the aforementioned limitations. Experiments on five UCI and ASU benchmark datasets demonstrate that GKFUAR significantly outperforms existing methods in terms of reduction efficiency, accuracy, and robustness. Furthermore, we deeply integrate GKFUAR with the TOPSIS decision-making framework to develop a new decision-making method named GKFUAR-TOPSIS, aimed at solving high-dimensional and complex FMADM problems under picture fuzzy set environments. Experimental results verify that the proposed model maintains high decision-making accuracy while achieving superior overall performance compared to several mainstream conventional methods.</p>