Preliminary Experiments on Adaptive RBF-Enhanced Fuzzy C-Means for Nonlinear Data Clustering

Fuzzy clustering methods, particularly Fuzzy C-Means (FCM), effectively handle uncertainty in cluster assignments but frequently encounter limitations when dealing with nonlinear and complex data patterns. This preliminary study introduces a novel adaptive distance-enhancement strategy, termed Radial Basis Function-enhanced Fuzzy C-Means (RBF-FCM), which integrates adaptively scaled Gaussian RBF kernels into conventional fuzzy clustering. Comparative experiments on six benchmark datasets, encompassing both nonlinear and structured scenarios, showed that the proposed approach delivered clustering results comparable to standard FCM and Kernelized FCM (KFCM) on simpler datasets, though encountering performance challenges in structured or noisy conditions. Further sensitivity analysis confirmed the critical role of the kernel-width parameter, highlighting its substantial effect on clustering quality. Although these initial experiments did not demonstrate decisive improvements over established methods, they revealed essential directions for future refinement, underscoring the necessity for systematic parameter optimization and adaptive strategies to realize the potential advantages of RBF-enhanced fuzzy clustering.