Reputation-Aware QoS Value Prediction of Web Services

Abstract

QoS value prediction of Web services is an important research issue for service recommendation, selection and composition. Collaborative Filtering (CF) is one of the most widely used methods which employs QoS values contributed by similar users to make predictions. Therefore, historical QoS values contributed by different users can have great impacts on prediction results. However, existing Web service QoS value prediction approaches did not take data credibility into consideration, which may impact the prediction accuracy. To address this problem, we propose a reputation-aware QoS value prediction approach, which first calculates the reputation of each user based on their contributed values, and then takes advantage of reputation-based ranking to exclude the values contributed by untrustworthy users. CF QoS prediction approach is finally used to predict the missing QoS values based on the purified dataset. Experimental results show that our approach has higher prediction accuracy than other approaches.