摘要
如何从海量的Mashup服务集中快速、准确的找到满足用户需求的Mashup服务,成为一个具有挑战性的问题.在M ashup服务发现中,预先对M ashup服务进行聚类,将大大缩小服务搜索的空间与范围,提高M ashup服务发现的效率与精度.本文提出一种新颖的融合K-Means与Agnes的Mashup服务聚类方法(MSCA).该方法,首先对Mashup服务中的Tag标签进行扩充和排序;其次,计算Mashup服务的集成相似性;接着,应用K-Means算法对Mashup服务相似度矩阵进行聚类,找到相似度较高的Mashup服务将其划分到N个原子簇中,再利用Agnes算法对N个原子簇进行层次聚类.最后,从Programmable Web上爬取了13082个Mashup服务作为实验对象,实验结果表明:相比传统的基于K-Means算法的Mashup服务聚类方法,MSCA方法的平均查准率和查全率分别提高了5.18%、5.84%,切实提高了服务聚类及发现的精度.
Howto rapidly and accurately select the users’ expected Mashup service has become a challenge problem. For Mashup service discovery,it will greatly reduce the space and scope of services searching to perform service clustering technology in advance,resulting in improving the efficiency and precision of Mashup service discovery. This paper proposes a novel Mashup Service Clustering Approach integrating K-Means and Agnes algorithms( MSCA). MSCA,first of all,will expand and rank the tag label of Mashup service. Secondly,it will calculate the Mashup service integration similarity. Thirdly,K-Means algorithm will be applied to clustering the Mashup service similarity matrix,and those Mashup services with the higher similarity will be found and divided them to N atom-clusters,and then Agnes algorithm will be used to performing hierarchical clustering to the N atom-clusters. Finally,13082 Mashup services are crawled from Programmable Web site and regarded as experimental dataset,and the experimental results showthat the average precision rate and recall rate of MSCA increased by 5. 18% and 5. 84% respectively,compared to the traditional Mashup Service Clustering Approach based on K-Means algorithm.
出处
《小型微型计算机系统》
CSCD
北大核心
2015年第11期2492-2497,共6页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(61402168
61402167
61272063)资助
软件工程国家重点实验室开放基金项目(SKLSE2014-10-10)资助
作者简介
黄兴,男,1988年生,硕士研究生,CCF会员,研究方向为服务计算;E-mail:jay1988528@163.com
刘小青,男,1962年生,博士,教授,博士生导师,研究方向为软件协同设计、服务计算等;
曹步清,男,1979年生,博士,副教授,研究方向为软件工程、服务计算与云计算;
唐明董,男,1978年生,博士,副教授,研究方向为网络科学,服务计算与云计算;
刘建勋,男,1970年生,博士,教授,博士生导师,研究方向为服务计算与云计算、工作流管理的理论与应用、大数据与商业智能等