For a particular clustering problems, selecting the best clustering method is a challenging problem.Research suggests that integrate the multiple clustering can improve the accuracy of clustering ensemble greatly. A new clustering ensemble approach based on the similarities in 2-mode networks is proposed in this paper. First of all, the data object and the initial clustering clusters transform into 2-mode networks, then using the similarities in 2-mode networks to calculate the similarity between different clusters iteratively to refine the adjacency matrix , K-means algorithm is finally used to get the final clustering, then obtain the final clustering results.The method effectively use the similarity between different clusters, example shows the feasibility of this method.
展开▼