Recommendation in an Evolving Service Ecosystem Based on Network Prediction
Keman Huang,Yushun Fan,Wei Tan
DOI: https://doi.org/10.1109/tase.2013.2297026
IF: 6.636
2014-01-01
IEEE Transactions on Automation Science and Engineering
Abstract:Service computing plays a critical role in business automation and we can observe a rapid increase of web services and their compositions nowadays. Web services, their compositions, providers, consumers, and other entities such as context information, collectively form an evolving service ecosystem. Many service recommendation methods have been proposed to facilitate the use of services. However, existing approaches are mostly based on all-time statistics of usage patterns, and overlook the temporal aspect, i.e., the evolution of the ecosystem. As a result, recommendation may consist of obsolete services and also does not reflect the latest trend in the ecosystem. In order to overcome this limitation, we propose an innovative three-phase network prediction approach (NPA) for evolution-aware recommendation. First, we introduce a network series model to formalize the evolution of the service ecosystem and then develop a network analysis method to study the usage pattern with a special focus on its temporal evolution. Afterward a novel service network prediction method based on rank aggregation is proposed to predict the evolution of the network. Finally, using the network prediction model, we present how to recommend potential compositions, top services and service chains, respectively. Experiments on the real-world ProgrammableWeb data set show that our method achieves a superior performance in service recommendation, compared with those that are agnostic to the evolution of a service ecosystem.Note to Practitioners-Understanding the usage pattern and the evolution mechanism can help better recommend services and their compositions to developers. Our hypotheses are: 1) historical information carries the usage patterns and evolution mechanism of the ecosystem and 2) services are not isolated but collaborative with each other, therefore, we should not only concern about individual services but also their correlations. Based on these hypotheses, our network prediction-based approach constructs the network series model based on historical information, and transform the evolution prediction into a network prediction problem. Furthermore, we predict the future behavior of the network based on link prediction. Based on the predicted network, we recommend potential compositions, services, and service chains that are evolution-aware and better reflect the up-to-date trend in the ecosystem.