Research On QoS Interval Based Web Service Selection

With the rapid development of Internet technology, software services industry has been going through a major revolution recently. The demand of software application is developinging thrive with the brand new computer technologies and software models, which shows a clear trend of development in service and networking. The software application model, i.e., Software as a Service(SaaS), makes it possible for users to customize software on demand. Now,more and more software developers or software companies deploy their private service on Internet. The service users can choose several satisfied services according to user own needs, and then customize new software services. By doing so, it greatly improves the flexibility for the customize services, and reduces the software development cycle, meanwhile it can save time and cost. With the number of services distributed on the Internet keeping growing, more and more Web services functions become the same or similar. However, their Quality of Service, QoS, mixed. It is very difficult to choose a satisfied and high-quality service from these functionally similar services for service users.In order to solve this problem, many scholars has launched several researches on Web service selection based on QoS, and then established many models. However, most of them, ignored the impact of the unstable external environment, do their researches on Web service selection with a single real value. Actually, QoS of Web services has unstability and dynamic characteristics or even fluctuate widely, which is very vulnerable to be influenced by user’s location, current network load and network state. Therefore, this paper focuses on considering the dynamic characteristics of Web services’ QoS during researching on Web services selection, and makes following contributions:(1) Considering the dynamic characteristics of Web services’ QoS, this paper proposes a method for selecting Web services based on uncertain QoS, which uses intervals to describe QoS values of services. Firstly, combining users QoS requirements with Web services’ interval QoS, and then using the normalized model and the benefit calculation model based on interval QoS to obtain the utility of every individual interval QoS. Secondly, synthesizing the weight of each QoS attribute presented by users to obtain the overall utility of each service, and then being sorted by its size. Finally, experimental results demonstrate that the proposed method is feasible and effective in selecting satisfactory services for users.(2) For Web services unknown QoS value, this paper proposed a prediction model to obtain an interval QoS. In the past prediction model, scholars get a certain real value. However, because of the QoS of Web services having dynamic characteristics in reality, service users tend to accept a predicted interval QoS, that is, an acceptable range of QoS value. Therefore, this paper proposed a series personalized prediction methods for Web service prediction based on interval QoS, including similar users based method, similar items based method and mixed based method. When user specified the interval length, the interval QoS, with greatest confidence, would be the predicted interval QoS for the missing value. Experimental results show that our personalized Web services prediction method based on interval QoS is more accurate and credible than other services collaborative filtering recommendation methods.