Research On Estimation And Prediction Of Confidence Interval For Web Service Composition QoS Based On Bootstrap

The QoS of Web Service Composition is not only a comprehensive indicator for judging the performance of Web Service Composition,but also an important basis for Web service selection.The attributes which is functional or non-functional of the Web Service Composition play an important role in the selection and recommendation of the Web Service Composition.The Web Service Composition that do not satisfy the QoS of the functional attributes or non-functional attributes cannot meet the user requirements well.So that estimation and prediction of QoS for Web Service Composition is the key to service selection and recommendation.At present,the research on estimating and predicting QoS of Web Service uses the point estimation method in statistics,but the QoS of Web Service Composition is always random due to factors such as the degree of network congestion and server load.The estimated QoS value of the same Web service combination in different time segments of the same place usually varies within a certain range.Therefore,the point estimation method estimates and predicts the QoS attribute value is inaccurate,which cannot truly reflect the objective law of QoS attribute value.Based on the QoS confidence interval estimation and prediction of Web services proposed by this research group,this paper uses the technique named Bootstrap of non-parametric statistics to further propose the estimation and prediction method of the QoS confidence interval of We Service Composition.The main innovations of the paper are as follows:(1)In order to calculate the QoS of Web Services Composition which has various structures by Bootstrap,it is proposed to use a nested list structure that is similar to Lisp language to represent Web service composition.The first element of list represents the structure type of the Web service composition(such as sequence,loop,parallel,choose),the subsequent elements of list can be either a Web service node that constitutes the Web Service Composition,or a new nested list that represents a nested structure of Web Service Composition.(2)It proposes an estimation method of QoS confidence interval of Web Service Composition on Bootstrap by using the historical QoS attribute value of the user to invoke the Web Service.(3)By using the Bootstrap of non-parametric statistical and collaborative filtering algorithm,the prediction method of QoS confidence interval for Web Service Composition is proposed based on the analysis of the similarity between users and the historical QoS of Web services invoked by similar users.Experiments were carried out on confidence interval estimation and prediction method of Web Service Composition QoS by the Web Service historical QoS in the WSDream dataset.The experiment estimates that each user in the dataset invokes the QoS confidence interval of the Web Service Composition and analyzes the experimental results briefly.The experiment randomly divides the users of dataset into 10% to 40% as the testset,and the remaining users in the dataset act as the trainingset.Predicting QoS confidence interval of Web Service Composition which testset user invoked by trainingset user.Comparing the actual call QoS attribute value of the test set user with the predicted QoS confidence interval,the accuracy of the prediction method is above 85%.