I. Introduction

Web services is a method that mutual exchange of data between software systems is available [1], in recent years, with the growing number of Web services, more and more web services are accessible. so how to choose a suitable service with the lowest cost has become a hotspot re-search. Owing to the characteristics of distributed system-s, system performance is highly depends on the relevant web services they used. Various types of applications put forward higher requirements in the performance of services, and the Web service QoS(Quality of Service, QoS) based on non-functional requirements drew the highest concern. Web service performance is highly dependent on QoS of the Web services they use, QoS is an important evaluation system in Web services quality, which represents the degree of reliability of a Web service, however the recent research is mainly to predict the accurate QoS values. In the field of service computing, the most frequently method that adopted to predict the web service QoS is collaborative filtering. The existing studies usually focus on the prediction of exact QoS value but never studies analyzes its fluctuation range, they only predict, evaluate and apply the QoS(mainly response time) as the point estimates when clients invoke the web service, but the Clients invoke the web service QoS(such as response time, throughput, blocking probability and systems resource utilization, etc.) are dynamic range, not a exactly value.