摘要
本文研究了p2p网络中基于内容的节点聚类。基于文件名关键词精确匹配的查询没有考虑文本语义及内容相似性。如果能够根据节点发布内容的相似性,建立节点聚类,信息查询在类内进行,必将提高查询效率。本文提出了一种基于增量学习的节点聚类方法,通过兴趣爬虫代理计算节点得分,据此判断一个节点是否可以加入节点簇。实验表明,节点簇的建立可以有效地提高 p2p 网络的查询效率。
This paper discusses the content-based peer clustering in peer-to-peer networks. Information retrieval based on accurate match of keywords in filenames ignores the document semantics and the similarity between documents. If peers are clustered according to the similarity between their released documents of a special interest topic, and the information query is executed among peers of a specific cluster, the efficiency should be improved. We propose an incremental learning approach to peer clustering, and employ an interest crawler agent to calculate a peer＇s score. Whether a peer joins in a cluster or not is determined by its score. Experimental results demonstrate that clustering of peers in hybrid p2p networks is both accurate and more efficient for information retrieval.
出处
《计算机科学》
CSCD
北大核心
2005年第12期184-187,共4页
Computer Science
关键词
对等网络
聚类
增量学习
Peer-to-peer networks, Cluster, Incremental learning
