改进Parzen窗解决高维数据聚类的方法研究  

Study of clustering using improved Parzen window on high-dimension dataset

作　　者：柴毅[1] 利节[1] 唐婧 

机构地区：[1]重庆大学自动化学院,重庆400030 [2]中国石油管道公司兰成渝输油分公司,成都610036

出　　处：《计算机工程与应用》2011年第8期135-137,共3页Computer Engineering and Applications

摘　　要：由于高维数据聚类的现实意义日益增强,而Parzen窗估计法仅对低维数据集聚类能获得良好的结果,随着维数增加,效率降低,因此对Parzen窗进行加权改进,通过多次仿真实验确定加权函数,将高维数据投射至低维空间,对其聚类,逐步投向高维空间,对结果矩阵进行优化处理,得到更为优良的聚类效果。Due to the realistic meaning of clustering on high-dimension dataset,excellent result can be acquired when clustering dataset using traditional Parzen window on low-dimension.With the higher dimension,the efficiency decreases quickly.Through many simulate experiments,the weighting function is gained.The high-dimension dataset is shadowed on lower-dimensional space,and clustered,then shadowed back to higher-dimensional space.The result matrix is optimized,and the better clustering effect is gained.

关 键 词：高维数据 PARZEN窗 聚类 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]