Mining and predicting of cold start service collaboration relationships in service systems

Services that have never been used and newly released services in Web service systems are called cold start services. A cold start service collaboration relationship mining and predicting method is developed to help service composition developers identify the characteristics of cold start services, increase attention to and usage of cold start services, and enhance the element diversity and robustness of service systems. The method first establishes a functional vector for each service using service description reconstruction and a functional topic analysis. Next, the method builds a collaborative vector for each non-cold start service based on its historical collaboration record and functional vector. Finally, the method compares the functional vectors of cold start services with the collaborative vectors of non-cold start services to predict the collaboration relationships for cold start services. Tests on real-world data show that this method more effectively predicts the relationships than state-of-the-art methods. © 2019, Tsinghua University Press. All right reserved.