摘要： ：针对传统分布式数据流挖掘算法的通信开销较大、分类精度较低的问题，提出一种基于支持向量数据描述的分布式数据流挖掘算法。利用局部站点快速更新数据流信息，采用支持向量机算法学习元级数据并传递到中心站点。中心站点负责接收及合并元级数据，形成全局分类结果。实验结果表明，该算法能在降低局部站点和中心站点网络通信量的同时，获得较高精度的全局分类结果。

关键词: 分布式数据流, 数据挖掘, 支持向量数据描述, 支持向量机, 增量式挖掘

Abstract: In distributed data stream mining, communication loads and global classification accuracy are main problems. In order to solve the problem, this paper presents a distributed data stream mining algorithm based on Support Vector Data Description(SVDD). Local site quickly updates data stream information, gets meta-level data by Support Vector Machine(SVM), and transmits them to central site. Central site receives and combines meta-level data, and learns global classification model. Experimental result shows that the algorithm can reduce transmission between local site and central site, and keep better classification accuracy.

Key words: distributed data stream, data mining, Support Vector Data Description(SVDD), Support Vector Machine(SVM), incremental mining

中图分类号: 

• TP391