Convergence analysis of semi-supervised clustering ensemble
Dahai Chen, Yan Zhu Yang, Hongjun Wang, Amjad Mahmood · 2013
Semi-supervised clustering ensemble fully integrates the advantages of semi-supervised learning, clustering analysis and ensemble learning, as well as improves the performance of clustering. There are many works on the algorithm and the consensus function of semi-supervised clustering ensemble, but there are few studies in the theoretical analysis. In this paper, we analyze the convergence of semi-supervised clustering ensemble, and propose a new relabeling approach for semi-supervised clustering ensemble by majority voting. We prove that semi-supervised clustering ensemble is able to boost weak learners to strong learners which can make very accurate predictions. The experimental results on standard data sets show that the semi-supervised clustering ensemble has better performance.