An Approach To Analyze The Confidence Interval Of Web Services QoS Based On Bootstrap

With the service-oriented architecture model is widely used,the Web service technology has also been considerable development.At present,the Web services are used more and more frequent on applications which in the market,Under the premise of ensuring the functional attributes of Web services,the non functional attributes of Web services QoS(Quality of Service)has become the focus attention for Web Users.The performance of Web services is highly dependent on the Web services QoS,QoS is an important evaluation system for Web services which represents the reliability degree of a Web service.At the moment,QoS prediction is the most critical step in service selection and service recommendation,and the existing research is mainly about the accurate prediction of QoS value,in the previous works,the QoS property such as response times observed by clients of Web services are predicated,evaluated and used as the point estimation.Actually,due to the random dynamic of the Internet and the Web server load,most of the QoS value is dynamic random and the client to invoke the Web service QoS(such as response time,throughput.etc)is the dynamic range,but not an exact value.In this paper with the help of non-parametric statistical Bootstrap technique,we proposed an approach to predict the confidence interval of Web services QoS,and accordingly the method to compare the performance of Web services.The contributions of this paper are as follows:1)Analyze the background of Web service technology and the related foun dation of statistics,and introduce the collaborative filtering algorithm,recomm endation system and the Qo S prediction algorithm based on collaborative filteri ng.2)Take use of the shapiro-wilk normal test method,we demonstrates that the Web service response times observed by clients are not normally distributed.3)With the kruskal-wallis hypothesis test,we demonstrates that the average response times experienced by clients invoke the same or different Web services are statistically significant difference.Therefore,we should estimate the confidence interval of the average response times observed by clients and analysis its volatilities.4)with the help of non-parametric statistical Bootstrap technique,we estimate and predict the confidence interval of user invoke the Web services' QoS,and thencompared performance of Web services,also a series of experiments to verify the feasibility and effectiveness of this method at last.The experimental results show that compared with the Web service Qo S a ccurate prediction,we can estimate and predict the confidence interval of Web services QoS based on Bootstrap method in this paaper,and it is feasible and efficient,it brings more fit to the actual evaluation standard for Web service s election and combination.