Abstract: Traditional clustering fusion method is to integrate all obtained cluster members, it doesn’t eliminate the inferior quality’ influence in  the integration completely, but it can have better advantages in theory from the selection and weight of clustering members to make clustering fusion, which select cluster members through fusioning any two clustering members instead of all clustering members once, and weight cluster members through the method of attribute weighting. It does experiment in the real and simulated datas. The result shows that the clustering algorithm can effectively deal with the difference of members quality, compared to the traditional clustering fusion method, it can get better results and have a good expansibility.

Key words: cluster fusion, cluster member selection, cluster member weighting

摘要： 传统的聚类融合方法通过融合所有成员实现融合，无法彻底消除劣质聚类成员对融合质量的影响，而从聚类成员的选择和加权两方面进行聚类融合，即先采用两两融合技术代替融合所有聚类结果进行聚类成员选择，然后进行基于属性的聚类成员加权，在理论上具有更好优越性。通过对真实数据和模拟数据的实验发现，该算法能有效处理聚类成员的质量差异，比传统聚类融合能得到更好的聚类结果，具有较好可扩展性。

关键词: 聚类融合, 聚类成员选择, 聚类成员加权