Two Partitional Methods for Interval-Valued Data Using Mahalanobis Distances

Two dynamic cluster methods for interval data are presented: the first method furnish a partition of the input data and a corresponding prototype (a vector of intervals) for each class by optimizing an adequacy criterion based on Mahalanobis distances between vectors of intervals and the second is an adaptive version of the first method. In order to show the usefulness of these methods, synthetic and real interval data sets considered. The synthetic interval data sets are obtained from quantitative data sets drawn according to bi-variate normal distributions. The adaptive method outperforms the non-adaptive one concerning the average behaviour of a cluster quality measure.