Joint learning of fuzzy embedded clustering and non-negative spectral clustering

Fuzzy k-means clustering is widely acknowledged for its remarkable performance in data clustering. However, its effectiveness must improve when dealing with high-dimensional data characterized by complex distributions, leading to subpar clustering results. To tackle this challenge, we introduce a novel clustering method named Joint Learning of Fuzzy Embedded Clustering and Non-negative Spectral Clustering (FECNSC). Initially, FECNSC utilizes rapid spectral embedding to reduce the dimensionality of the data. Subsequently, it incorporates fuzzy clustering and non-negative spectral clustering in a unified framework. The novel fuzzy clustering method enhances fuzzy membership by regularising of non-negative spectral clustering. Our experimental results demonstrate the overall superiority of FECNSC in terms of accuracy, normalized mutual information, and purity across various benchmark datasets, surpassing multiple advanced methods. Therefore, FECNSC is an efficient solution for managing data with complex distributions.