基于DBSCAN的最优密度文本聚类算法  

Optimal density text clustering algorithm based on DBSCAN

作　　者：李群[1] 袁津生[1] 

机构地区：[1]北京林业大学信息学院,北京100083

出　　处：《计算机工程与设计》2012年第4期1409-1413,共5页Computer Engineering and Design

摘　　要：为有效地弥补全文搜索引擎的不足,提出了一种动态求解的最优密度聚类算法并加以实现。该算法构造了一颗簇关系树,将两种典型聚类算法:密度聚类算法DBSCAN和层次聚类算法BIRCH进行有效结合,对聚类参数ε进行动态求解,以达到参数ε的最优。与其它文本聚类算法相比,该算法的查询结果与用户感兴趣的主题相关度较大,对具有二义性的关键词有较高的查准率,能有效提升搜索引擎的查询效率,加快用户搜索信息的速度。To make up for lack of full-text search engine effectively, a dynamic solution for optimal density clustering algorithm is proposed and designed. The algorithm uses a combined regimen of density clustering algorithm and hierarchical clustering algo- rithm, and between a cluster tree is constructed, at the same time a dynamic solution for clustering parameters ~ is used to achieve the optimal parameter ~. Compared with other text clustering algorithms, the algorithm can return the results which related to the user＇s interest topics, improve the query accuracy of the ambiguous words, effectively improve the efficiency of search engine, contribute to quick search for information.

关 键 词：搜索引擎 文本聚类 密度聚类 簇关系树 动态聚类 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]