<p>This study proposes fuzzy support vector machines (FSVM) with a novel multi-class framework termed iterative multi-criteria classification (IMCC) and introduces several normalized-distance reciprocal (NDR) membership functions. The framework is applied to sovereign credit rating prediction, a complex task influenced by economic, financial, and geopolitical factors. To address the limitations of traditional multi-class methods, we develop four approaches: Sequential removal classification (SRC), Reordered SRC (RSRC), Max-probability reassignment classification (MPRC), and Reordered MPRC (RMPRC). These methods iteratively refine class assignments or reassign ambiguous instances, explicitly modeling inter-class relationships and overlapping boundaries. Experimental evaluation on financial and credit rating datasets demonstrates that the proposed methods outperform conventional multi-class schemes such as One-vs-All, One-vs-one, and error-correcting output codes in terms of accuracy, precision, and recall. The integration of NDR-based membership functions further improves robustness by down-weighting noisy and ambiguous observations, leading to more stable and interpretable decision boundaries. Overall, this study establishes an IMCC-based FSVM framework with NDR membership functions that jointly enhances robustness, interpretability, and multi-class discrimination in sovereign credit rating prediction. The proposed framework remains applicable to broader financial and ordinal multi-class classification problems.</p>