Entropy in Fuzzy k-Means Algorithm for Multi-view Data

Abstract

Multi-view data clustering plays a crucial role in various real-world applications. This kind of data from various domains can exhibit a range of distributions, making it challenging for algorithms to uncover robust patterns. This paper extends the fuzzy k-means clustering algorithm to cluster multi-view data. The objective function includes two additional matrixes to measure the compactness of each view and the importance of individual features. The objective function also includes entropy weights. Experiments on real-life data indicate that the proposed algorithm outperforms current state-of-the-art algorithms. These set of algorithms comprises of clustering techniques that incorporate variable weighting, such as W-k-means [11], LAC [9], and EWKM [13], along with a multiview clustering algorithm called TW-k-means [6]. The evaluation of the algorithms involves measuring their accuracy, as well as comparing their respective running times. A comprehensive discussion on the proposed algorithm’s properties was conducted, where all its parameters were fine-tuned and analyzed in detail.