Research On The QoS Prediction For Web Services In Cloud Computing

With the rapid development of cloud computing and Web services technology, software development paradigm has a great and corresponding change. The traditional development mode, identified by such characteristics as multiple usages for one developed version, has also gradually transformed to the new paradigm of "software as a service", on-demand providing, virtualization in cloud and etc. As a new kind of open business providing mode as well as loose coupling and reusable style, Web services have become the important resources for service providing in cloud center. Also, they are the basic components for building software business processes and applications. In recent years, the number of Web services deployed in cloud center or network is explosively increasing. As a result, quite a few Web services have the identical or similar functions. After the functional requirements are satisfied, users will pay continuous concern on the quality attributes of Web services (QoS for short here). Consequently, how to select the most suitable services w.r.t. QoS for end-client users is a key and challenging research topic.Recently, QoS prediction technology has been viewed as an effective way to assist Web services recommendation. In this paper, problems on the QoS value prediction and rankings prediction have been deeply investigated. For QoS value prediction, we attempt to settle the problem through combining Pearson similarity with Slope One method together, and proposed a new prediction algorithm SASO. In such algorithm, the service invocation information from the top-k nearest neighbor users is fully utilized for predicting the missed QoS values. At the same time, the SPC-based smoothing technique is used to reduce the interference from noise data. In order to verify the feasibility and effectiveness of our SASO algorithm, some comparative experiments are performed on the real-word data set. The results show that SASO outperforms the existing methods like WSRec and basic Slope One for QoS value prediction problem. With regard to QoS ranking, a corresponding prediction algorithm named PSORank is presented here. The main components of our algorithm include the construction of fitness function, similar user finding, and the design of optimization search strategies. More importantly, the priority probability of service, the consistence between the current user and the whole group are also taken into consideration. Similarly, the experiments are also performed so as to evaluate the effectiveness of PSORank algorithm. The experimental results reveal the facts that PSORank algorithm is very suitable to QoS ranking prediction problem, and also can overcome the sensitivity problem of population in evolutionary search. It’s not hard to find that, our research on QoS prediction of Web service can provide some guidelines for Web services recommendation, and also can effectively improve users’ satisfaction for Web service selection.