摘要
针对聚类融合问题,考虑了聚类成员的质量和噪声对聚类结果的影响,提出一种加权迭代的聚类融合模型,利用粗糙集理论中的决策表属性重要性的信息熵来衡量聚类成员的重要性,迭代更新聚类成员的权重。该文在模拟和真实数据集上进行了校验。结果表明,该模型能较好地处理聚类成员间的质量差异,并能有效地消减噪声对融合的影响,从而得到更好的聚类融合结果。
An iterative weighted cluster ensemble (IWCE) model was developed taking into account the qualities of the cluster members and the noise in the cluster ensemble. The model evaluates the significance of each cluster member using information measuring the attribute significance in the rough set and iteratively updates the weight values. Experiments on several synthetic and real data sets show that the model can handle different-quality cluster members and effectively lessens the effect of noise. Therefore, the model provides better ensemble results than general cluster ensemble methods.
出处
《清华大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2009年第8期1106-1108,1121,共4页
Journal of Tsinghua University(Science and Technology)
关键词
聚类融合
共生矩阵
信息熵
加权迭代模型
cluster ensemble
co-association matrix
information entropy
iterative weighted model