一种不确定数据流子空间聚类算法

提出了基于滑动窗口的不确定数据流子空间聚类算法USSC,它应用采样时加权值的方法来选择初始化聚类中心点,采用滑动窗口SW缓存一段时间的元组作为聚类对象,并提出一种新的离群点处理机制来排除离群点Opo为适应不确定数据流元组不确定特性,该算法使用基于隶属度的非分割聚类方法来确定一个元组只能划分到一个簇中.试验结果表明,USSC算法与同类型的算法相比有较好的聚类效果和较快的聚类速度,而且其自身拥有很强的可伸缩性.

Clustering algorithm over uncertain data streams based on sliding window

The paper proposes a new clustering algorithm USSC for solving clustering problem on uncertain data stream, which uses sliding window to save tuple in a period of time as clustering objects. The character of uncertain data stream is that the tuples in it is no ture. In order to adapt this character, the algorithm uses weighted values to define sub-space cluster center point. In addition, it presents a new outliers mechanism to remove outliers. The experiments show that the SWCUS algorithm has better clustering effect and faster clustering speed than other algorithms as well as better scalability.