A Novel Clusterer Ensemble Algorithm Based on Dynamic Cooperation
Kai Kang, Huaxiang Zhang, Ying Fan · 2008
As the better generalization ability of clusterer ensemble methods, they are widely applied to diverse domains. But now many challenges still exist. One of the drawbacks of the ensemble is, ignoring the valuable information contained in the process of training component clusterers. This paper explores a new ensemble method for cluster analysis based on dynamic cooperation, and this method adjusts the centroids using the information provided by all component clusterers. The basic idea is to make training information fully sharable in the ensemble method. We apply the proposed ensemble method to the UCI benchmark data sets and the experimental results show that the approach provides a practical solution.