考虑数据排序的改进CABOSFV聚类  

Improved CABOSFV clustering considering data sort

作　　者：武森[1] 王静[1] 谭一松[1] 

机构地区：[1]北京科技大学经济管理学院,北京100083

出　　处：《计算机工程与应用》2011年第34期127-129,共3页Computer Engineering and Applications

基　　金：国家自然科学基金(No.70771007);中央高校基本科研业务费专项资金资助(No.FRF-TP-10-006B)~~

摘　　要：CABOSFV是基于稀疏特征进行高维数据聚类的高效算法,但算法的聚类质量受数据输入顺序的影响。针对此问题,提出考虑数据排序的改进CABOSFV聚类(CABOSFV_CS),通过定义稀疏性指数来描述数据的稀疏特征,并按照稀疏性指数升序对数据进行排序以改进CABOSFV算法的聚类质量。采用UCI基准数据集进行实验,结果表明与传统的CABOSFV算法相比,CABOSFV_CS有效地提高了聚类准确率。CABOSFV is an efficient algorithm based on sparse feature for high dimensional data clustering.However the clustering quality of the algorithm is sensitive to the order of input data.To this problem,improved CABOSFV clustering considering data sort（CABOSFV_CS） is proposed,which describes the sparse feature of data by defining a new concept sparseness index and improves the clustering quality of CABOSFV by sorting data according to the ascending sequence of sparseness index.UCI benchmark data sets are used to compare CABOSFV_CS with traditional CABOSFV algorithm.The empirical tests show that CABOSFV_CS increases the clustering accuracy effectively.

关 键 词：CABOSFV算法 高维数据 稀疏特征 聚类 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]