The fuzzy C-means clustering algorithm used to process the high-dimensional datasets has the problems of high computational complexity,poor algorithm generalization ability and low calculation accuracy. Considering the difference of fea-ture attribute for clustering contribution,a new reduction algorithm based on attribute importance is proposed on the basis of the thought of multiattribute fuzzy C-means clustering. In order to verify its validity,the comparative analysis was performed in UCI datasets for the proposed algorithm,factor analysis method and reduction method based on rough set theory. The experimental results show this method has wider application range,and better performance on the datasets whose average standard deviation is large or the inter-class centre distance is far.%模糊C均值聚类算法在处理高维数据集时,存在计算复杂度高,算法泛化能力差,计算精度低等问题.考虑到特征属性对聚类的贡献程度的差异,在多属性模糊C均值聚类的思想上,提出一种基于属性重要性的约简算法.为验证有效性,在UCI数据集上,将新算法与因子分析法和粗糙集理论约简方法进行比较分析.实验结果表明,该方法具有更好的泛用性,在平均标准差大或类间中心距离较远的数据集上具有更好的性能.
展开▼
