A Comparative Study on Fuzzy Clustering Algorithms in Medical Data Analysis

Abstract

 Fuzzy clustering algorithms are rapidly applied in medical data analysis because they can capture uncertainty and overlapping patterns more effectively than traditional hard clustering methods. This paper presents a comparative study of five widely studied approaches such as Conventional Fuzzy C-Means, Kernelized Fuzzy C-Means, Fuzzy Possibilistic C-Means, Gaussian Kernel-based Intuitionistic Fuzzy C-Means, and  Fuzzy Gustafson–Kessel algorithm. This work examines theoretical foundations, strengths, and limitations of the above said methods, specifically emphasizing applications in various cancer datasets including lung, breast, and cervical cancer. Our findings suggest that while the classical Fuzzy C-Means remains reliable for linearly separable and relatively clean datasets, kernel-based and covariance-adaptive methods offer advantages for nonlinear and anisotropic patterns, and possibilistic models improve robustness to noise and outliers. Collectively, these approaches demonstrate strong potential for supporting early diagnosis, patient stratification, and personalized treatment planning in oncology.