A robust fuzzy \(k\)-means clustering model for interval valued data (Q880908): Difference between revisions

The authors propose a \(k\)-means clustering algorithm for multivariate interval-valued data. It generates fuzzy clusters which are represented by the membership degree matrix \(\{u_{iq}\}\), where \(u_{iq}\) indicates the membership degree of the \(i\)-th observation to the \(q\)-th cluster (\(q=1,\dots,k\)). The noise approach is used to reduce the influence of the outliers on the whole partition, i.e., the outliers are assigned to a special cluster of data. The degree of membership to this noise cluster is characterized by \(u_{i*}=1-\sum_{q=1}^k u_{iq}\). Results of simulations and application to a ``Fat and Oils'' data set are presented.