A Dual-Weighted Gaussian Kernel-Based Method for Fuzzy Co-Clustering
José Nataniel A. de Sá, Marcelo R.P. Ferreira, Francisco de A.T. de Carvalho · 2025
Unlike traditional clustering methods that seek groups of objects or variables separately, co-clustering algorithms simultaneously group objects (rows) and variables (columns), aiming to identify homogeneous data blocks. Therefore, not only the variables can represent noise, but also the objects. In this article, we propose a Dual Weighted Gaussian Kernel Fuzzy Double K-means (DWGKFDK), a fuzzy co-clustering model that automatically computes the weights of both objects and variables, making it suitable for scenarios with not only irrelevant variables but also irrelevant objects. The DWGKFDK algorithm reduces the negative impact of these objects and variables by assigning lower weights for them. To handle non-linear data, we use the Gaussian kernel, which is commonly employed for grouping non-linearly separable data. Experiments on synthetic and real datasets confirmed the efficiency of the new model compared to four state-of-the-art fuzzy co-clustering algorithms.