摘要
针对多数据源或异构数据集,采用单个核函数的聚类效果不理想的问题,以及考虑到不同属性对不同类别重要性的差异,本文提出了一种属性加权多核模糊聚类算法(WMKFCM)。该算法将多核模糊聚类算法与属性加权核模糊聚类算法相结合,不仅能够处理单个核函数不能满足待聚类数据集聚类准确度要求的问题,而且能在聚类过程中根据不同类的具体特性动态调整各个属性对于不同类别的重要性。聚类实验表明,在牺牲一定的运行时间和迭代次数的前提下,相比于属性加权核模糊聚类算法和多核模糊聚类算法,属性加权多核模糊聚类算法具有更高的聚类准确度。
Considering the problem of unsatisfactory clustering effect of single kernel function for multiple data sources or heterogeneous data sets, and taking into account the difference in importance of different attributes for different categories, this paper proposes an attribute-weighted multikernel fuzzy clustering algorithm (AWMKFCM)). The algorithm combines the multi-kernel fuzzy clustering algorithm with the attribute-weighted single-kernal fuzzy clustering algorithm. It not only can handle the problem that the single kernel function can not meet the clustering accuracy requirements of the clustering data set, but also can adjust the importance of each attribute dynamically to different categories according to the specific characteristics of different types in the clustering process. Clustering experiments show that the attribute-weighted multi-kernal fuzzy clustering algorithm has higher clustering accuracy than the attribute-weighted single-kernel fuzzy clustering algorithm and multi-kernal fuzzy clustering algorithm under the premise of a certain amount of running time and iterations.
作者
阚云
包振强
张照岳
Yun Kan;Zhenqiang Bao;Zhaoyue Zhang(College of Imformation Engineering, Yangzhou University, Yangzhou Jiangsu)
出处
《计算机科学与应用》
2018年第6期961-969,共9页
Computer Science and Application
关键词
模糊聚类
混合核函数
属性加权
Fuzzy Clustering
Multi-Kernal
Attribute-Weighted