Collaborative Web Service QoS Prediction on Unbalanced Data Distribution

Abstract

QoS prediction is critical to Web service selection and recommendation. This paper proposes a collaborative approach to quality-of-service (QoS) prediction of web services on unbalanced data distribution by utilizing the past usage history of service users. It avoids expensive and time-consuming web service invocations. There existed several methods which search top-k similar users or services in predicting QoS values of Web services, but they did not consider unbalanced data distribution. Then, we improve existed methods in similar neighbors' selection by sampling importance resampling. To validate our approach, large-scale experiments are conducted based on a real-world Web service dataset, WSDream. The results show that our proposed approach achieves higher prediction accuracy than other approaches.