New clustering methods for interval data (Q880906)

Two new clustering methods are proposed for multivariate interval data. They are modifications of the dynamic \(k\)-clustering algorithm based on minimization of distances from elements of the cluster to its prototype (centre). The number \(k\) of clusters is fixed. A sum of Hausdorff distances between intervals is used in the first modification and a specific dissimilarity measure based on a discretization of the intervals is used in the second modification. These algorithms are applied to data on monthly minimal and maximal daily average temperatures in 60 meteorological stations in China.