基于概要数据结构的高维数据流聚类算法

High-dimensional Data Streams Clustering Algorithm Based on Synopsis Data Structure

为了在高维数据流中有效地形成聚类,针对经典算法CELL-Tree存在的问题,提出一种新的概要数据结构PL-Tree以及基于此数据结构的算法PLStream,并采取衰减窗口模式来适应数据流的变化,采用剪枝策略控制内存中聚类模型的规模.实验表明,PLStream算法能较好地适应高维数据流,比CELL-Tree算法具有更好的时间和空间效率.

To form clustering effectively in the high-dimension data streams, focusing on the questions in the classical algorithm CELL-Tree, a new synopsis data structure PL-Tree and the algorithm PLStreem based on it are put forward. And the damped window model is adopted to adjust to the changes of the data streams. The cluster model＇s scale of the internal storage is controlled by pruning methods. The research suggests that the PLStream algorithm can adjust to the high-dimensional data stream better and is superior to the CELL-Tree algorithm in better temporal and spatial efficiency.