摘要
针对已有的基于特征加权距离的双指数模糊子空间聚类算法(DI-FSC)没有充分利用数据集中的已知信息,提出利用少量监督信息辅助聚类过程的特征加权距离的半监督模糊子空间聚类算法(SS-FSC).通过在该算法中加入特征加权距离来改善传统聚类中利用欧氏距离计算数据点之间差异的不足;同时向约束条件中引入指数r和β,增加了算法的灵活性.实验表明,所提出的算法在少量监督信息的辅助下,在真实数据集上有较好的聚类效果.
In view of the existing double-indices fuzzy subspace clustering algorithm based on feature weighted distance （ DI-FSC ） doesn＇t make full use of the known information of the datasets ,put forward a new kind of clustering method using a small amount of supervision information to assist the process of clustering. By introducing the feature weighted distance to improve the disadvantage of using the Euclidean metric to compute the distance between data points. At the same time,adding the index andto the constraint condi- tions to increase the flexibility of the algorithm. Experimental results show that the proposed algorithm under little supervision information on real datasets has better clustering effect.
出处
《小型微型计算机系统》
CSCD
北大核心
2017年第2期405-410,共6页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(61272210)资助
关键词
半监督
模糊聚类
子空间
特征加权距离
semi-supervised
fuzzy clustering
subspace
feature weighted distance
