中文摘要：

Web服务及SOA技术的出现为Web应用架构提供了一个新的范式．未来会有大量功能相同或相近的Web服务，如何从中优选出符合用户需要的Web服务已成为一个正在研究的问题．在研究了现有的Web服务组中Web服务优选方法基础上，针对其仅局限于功能优选的不足，提出了一种面向QoS的Web服务组中Web服务优选方法．在定义Web服务QOS向量特征分量、QoS向量、最大相似度、QoS测试预言、QoS向量特征分量测试预言等参数基础上，基于层次聚类思想实现QoS向量聚类，依据最大相似度控制聚类层次，之后利用QoS测试预言、QoS向量特征分量测试预言及决策树实现优选．实验结果表明该方法是有效的，克服了以前的方法仅限于功能优选的局限性．

英文摘要：

The emerging technologies of Web services and SOA have supplied a new paradigm for Web application architecture. Because there will be a lot of Web services in the future, which have the similar or same functions, the selection for Web services under group testing has become an ongoing research topic. Based on the related work of Web service selection under group testing, the authors propose a QoS-oriented approach for Web service selection under group testing to overcome the localization of the existing methods, which are only based on function-optimized selection for Web services under group testing. By defining QoS vector, its characters and maximum similar degree, QoS testing oracle, QoS character testing oracles and so on, QoS vectors are clustered by using hierarchical clustering. This approach adjusts the clustering hierarchy dynamically according to maximum similar degree. After the clustering is finished, the selection for Web services under group testing is made by use of QoS testing oracle, QoS character testing oracles and decision tree. It is proved by experiment results that the approach is effective and able to overcome the localization of the existing methods which are only based on function-optimized selection for Web services under group testing.