Collaborative QoS Prediction via Matrix Factorization and Topic Model

With the explosive growth of Web services on the Internet, Quality-of-Service-based (QoS) service selection is becoming an important issue of service-oriented computing. The QoS values of services to current users are all supposed to be known in the previous works, while lots of them are not known in reality. In order to predict the missing data, many approaches have been employed in recent years. However, those approaches don't carefully consider the online cold-start scenario where many new registered Web services haven't been involved even once. This paper proposes a collaborative QoS prediction framework named CQP integrating matrix factorization with probabilistic topic model. This approach builds an integral latent user and Web service representative space, and can be applied online to predict QoS value and handle the online cold-start problem. To validate our methods, some approaches and our algorithm are conducted on a real-world dataset. The experiment result demonstrates that the proposed approach outperforms the previous works in prediction accuracy.