摘要
现有的加权模糊C均值聚类算法中,属性加权是一个不断迭代、重复计算的过程,费时费力。针对这种情况,提出Fisher线性判别率进行属性加权。算法首先直接计算每一维属性对模糊聚类的贡献度,其次对所有属性的贡献度进行归一化处理然后加权聚类。在人工和实际数据集所做实验表明:该算法在提高聚类速度的同时,聚类效果上也优于其他同类加权模糊C均值聚类算法。
In existing weighted fuzzy C-means clustering algorithm,attribute weighting is a process of ongoing iteration and repeated calculation,which is time-consuming and laborious. In view of this situation,we proposed to use Fisher linear discriminant rate in attribute weighting. First,the algorithm calculates directly the contribution degree of every dimension attribute on fuzzy clustering; secondly it makes the normalisation processing on the contribution degrees of all attributes followed by weighted clustering. The experiments carried out on artificial and real datasets show that while improving the clustering speed,the method has the superiority over other similar weighted fuzzy C-means clustering algorithm in terms of clustering effect.
出处
《计算机应用与软件》
CSCD
2016年第7期273-277,共5页
Computer Applications and Software
基金
吕梁学院校内自然科学基金项目(zrxn201308)
关键词
模糊
C
均值聚类
FISHER
线性判别率
属性加权
分配系数划分
熵
隶属度
彩色图像分割
颜色空间
Fuzzy C-means clustering
Fisher’s linear discriminant
Attribute weighting
Distribution coefficient partition
Entropy Degree of membership
Colour image segmentation
Colour spaces