Abstract
When dealing with multiple clustering solutions, the problem of extrapolating a small number of good different solutions becomes crucial. This problem is faced by the so called Meta Clustering [12], that produces clusters of clustering solutions. Often such groups, called meta-clusters, represent alternative ways of grouping the original data. The next step is to construct a clustering which represents a chosen meta-cluster. In this work, starting from a population of solutions, we build meta-clusters by hierarchical agglomerative approach with respect to an entropy-based similarity measure. The selection of the threshold value is controlled by the user through interactive visualizations. When the meta-cluster is selected, the representative clustering is constructed following two different consensus approaches. The process is illustrated through a synthetic dataset.