基于相似性分析的软件构件聚类研究  

Research on Clustering Software Components Based on Similarity Analysis

作　　者：陈德华[1] 韩忠明[1] 乐嘉锦[1] 

机构地区：[1]东华大学计算机学院,上海200051

出　　处：《小型微型计算机系统》2005年第12期2207-2211,共5页Journal of Chinese Computer Systems

基　　金：国家"八六三"高科技研究发展计划项目(2002AA423430)资助

摘　　要：设计了一种对基于XM L描述的软件构件进行聚类的算法(即基于模拟退火的构件聚类算法),该算法通过模拟金属退火基本原理对构件库中的软件构件聚类进行全局优化.构件聚类时,根据一般意义的树间编辑距离,提出一种用于判断基于XM L描述的构件间是否相似的度量测度(称为XM L编辑距离).利用XM L编辑距离,可将构件间相似性度量的时间复杂度限制在多项式级,且能保持构件的XM L描述文档的节点语义信息和节点间的祖孙嵌套关系.最后,在构件库测试模型上进行实验,结果证实了基于模拟退火的构件聚类算法在构件查询实践中的可行性和有效性.This paper proposes a native algorithm for clustering software components described by XML.. simulated annealingbased clustering algorithm of software component. The aim of the algorithm is to achieve globally-optimized component clusters in component repository by simulating the basic principles of metal anneal. During clustering, a quantitative approach is presented in this paper to measure the similarity between two components described by XML, called XML edit distance. With XML edit distance, the computation of the similarity between components is resolved in polynomial time, furthermore, the semantic information of nodes and the relationship of ancestor-descendent among nodes in XML documents can also be preserved. Finally, the results of experiments on a testing system of component repository confirm the feasibility and efficiency of simulated annealing-based clustering algorithm in component retrieval.

关 键 词：软件构件 XML描述文档 XED 聚类算法 模拟退火算法 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]