Existing Web service selection approaches usually assume that preferences of users have been provided in a quantitative form by users. However, due to the subjectivity and vagueness of preferences, it may be impractical for users to specify quantitative and exact preferences. Moreover, due to that Quality of Service (QoS) attributes are often interrelated, existing Web service selection approaches which employ weighted summation of QoS attribute values to compute the overall QoS of Web services may produce inaccurate results, since they do not take correlations among QoS attributes into account. To resolve these problems, a Web service selection framework considering user's preference priority is proposed, which incorporates a searching mechanism with QoS range setting to identify services satisfying the user's QoS constraints. With the identified service candidates, based on the idea of Principal Component Analysis (PCA), an algorithm of Web service selection named PCA-WSS (Web Service Selection based on PCA) is proposed, which can eliminate the correlations among QoS attributes and compute the overall QoS of Web services accurately. After computing the overall QoS for each service, the algorithm ranks the Web service candidates based on their overall QoS and recommends services with top QoS values to users. Finally, the effectiveness and feasibility of our approach are validated by experiments, i.e. the selected Web service by our approach is given high average evaluation than other ones by users and the time cost of PCA-WSS algorithm is not affected acutely by the number of service candidates.
在大量相似Web服务共存竞争的环境下,基于服务质量的Web服务选择成为服务计算领域的热点问题之一.现有的Web服务选择方法主要研究单个服务请求或多个合作关系的服务请求共同选择某一个服务的情形,未考虑多个独立的服务请求同时请求同种功能服务的互相竞争性.针对该问题,根据Web服务与服务需求之间的匹配度,利用0-1整数规划建立全局优化服务选择模型,并结合实际提出通用可行的解决多请求的全局优化服务选择算法(global optimal service selection for multiplerequests,GOSSMR),在保证Web服务需求质量得到满足的情况下,避免过多的请求同时选择同一个服务,做到资源合理利用,避免服务负载失衡,提高系统的性能.仿真实验验证了模型算法的可行性和有效性.
