基于Fisher线性判别率的加权K-means聚类算法

Weighted K-means clustering algorithm based on Fisher's linear discriminant ratio

为提高K-means聚类效果,采用Fisher线性判别率的方法确定特征在聚类中的贡献度并依此对特征进行加权聚类。在人工和实际数据集上所做的实验表明,本方法在聚类效果上优于其他同类加权K-means聚类算法。

To improve the clustering effect of K-means,the features were successively weighted by using the method of Fisher＇s linear discriminant ratio according to their contribution in clustering and the data were clustered. The experimental results done on synthetic and real data show that the method has superiority over other weighted K-means clustering algorithm in the clustering effect.