Fuzzy clustering, Credibilistic fuzzy c-means, Spherical space, Gene expression, Cancer data

This article presents a robust noise-resistant fuzzy-based algorithm for cancer class detection. High-throughput microarray technologies facilitate the generation of large-scale expression data; this data captures enough information to build classifiers to understand the molecular basis of a disease. The proposed approach built on the Credibilistic Fuzzy C-Means (CFCM) algorithm partitions data restricted to a p-dimensional unit hypersphere. CFCM was introduced to address the AQ2 noise sensitiveness of fuzzy-based procedures, but it is unstable and fails to capture local non-linear interactions. The introduced approach addresses these shortcomings. The experimental findings in this article focus on cancer expression datasets. The performance of the proposed approach is assessed with both internal and external measures. The fuzzy-based learning algorithms Fuzzy C-Means (FCM) and Hyperspherical Fuzzy C-Means (HFCM) are used for comparative analysis. The experimental findings indicate that the proposed approach can be used as a plausible tool for clustering cancer expression data.