基于聚类准则融合的加权聚类集成算法

摘要

聚类集成是聚类分析中的一个重要技术手段,能有效地提高聚类结果的准确性、鲁棒性和稳定性.利用现有的聚类准则提出了一个新的评价指标,用于基聚类结果的有效性评估,并把评估值作为基聚类结果的权重来进行加权聚类集成.在UCI真实数据集上对提出的基于聚类准则融合的加权聚类集成算法进行了测试,实验表明新提出的算法比已有的集成聚类算法具有更高的准确率和回收率,可以得到更好的集成聚类结果.%Clustering ensemble is an important technique in cluster analysis,which can effectively improve the accuracy,robustness and stability of clustering results.A new clustering criterion is proposed based on the fusion of existing clustering indexes.The new clustering criterion is used to evaluate the validity of the basic clustering results,and the results of the evaluation are used as the weights of the basic clustering resuits.The weighted clustering ensemble algorithm based on clustering criterion is tested on the UCI dataset.The experiments results on multiple real data sets suggest that the proposed method almost always outperforms well-known cluster ensemble techniques.