Combining hierarchical clusterings using min-transitive closure

In the past, clusterings combination approaches are based on “flat” clustering algorithms, i.e. algorithms that operate on non-hierarchical clustering schemes. These approaches, once applied to a hierarchical clusterings combination problem, are not capable of taking advantage of the information inherent in the input clusterings hierarchy, and may thus be suboptimal. In this paper, a new hierarchical clusterings combination framework is proposed for combining multiple dendrograms directly. In this framework, the description matrices of the primary hierarchical clusterings are aggregated into a transitive consensus matrix with which the final clustering is formed. Experiments on real-world datasets indicate that this framework provides solutions of improved quality.

• Type: Conference
• Year: 2008
• Where: ICPR
• Authors: Abdolreza Mirzaei, Mohammad Rahmati
• Keywords: Clusterings Combination Approaches, Computer Vision, Hierarchical Clusterings Combination, ICPR 2008, Non-hierarchical Clustering Schemes