Research On Web Service QoS Prediction Based On Time Series Analysis

With the development of the Internet,the traditional software model has been unable to meet the needs of complex businesses.More and more companies have turned to Service-Oriented Architecture(SOA).Web services have become the standard for implementing SOA architecture.Complex business functions are achieved by combining multiple web services.In this case,any problem with a Web service will lead to problems in the operation of the entire service system.People are generally aware of the importance of Quality of Service(QoS).For the system to operate stably in a highly dynamic distributed environment,the prediction of QoS has become a hot issue in service comp uting.The prediction method based on time series analysis has been widely used.The first problem faced in the QoS prediction applied to Web services is the missing value problem.The average user does not have continuous access to a service at all times,and does not have a service for all services.Call records;second is the accuracy of prediction,QoS data is highly volatile,unlike traditional software reliability models that can obtain interpretable parameters.Aiming at these two problems,this pape r proposes a method for estimating missing values based on similarity in time series.By mining the similarity of time series,aggregating similar time series and estimating missing values,the effect of QoS data sparse on forecasting is effectively reduce d.Aiming at the characteristics of high volatility of QoS data,the traditional model is combined with Kalman filtering method to improve the fluctuating response sensitivity to QoS of Web services.For the low applicability of the traditional time series model,the use of recurrent neurons with memory cells is proposed.The network model predicts the QoS of the Web service,improves the accuracy of the prediction,and analyzes the influence of various factors on the accuracy of the prediction.The methods proposed in this paper all have multiple sets of experimental verification on the public data set.The verification results show the effectiveness and advantages of the proposed method.