Clustering Ensemble Based on Hierarchical Partition

Many clustering ensemble algorithms need to predesign initial thresholds before partition data points, which is supervised learning and directly influence the efficiency of clustering. In order to cluster data points under fully unsupervised situation, the hierarchical partition is introduced in this paper. The proposed algorithm makes use of the distribution of results of all clustering memberships by constructing the m-subset of Descartes with the support degree. The theorems and definitions advanced in this paper are detailed proved. Finally, the proposed algorithm is applied in practice and results show that it is effective.