Integrating User Invocation Data and Extended Semantics for Service Community Discovery

We present in this paper an integrated service discovery framework based on Non-negative Matrix Factorization (NMF). NMF provides an effective means to cluster high-dimensional sparse data with both high clustering accuracy and good interpretability of the clustering result. This makes NMF especially suitable for service community discovery by clustering the Web service description data. Nevertheless, as the standard service description language, WSDL, rarely offers rich service description, accurate discovery of service communities remains as a central challenge. The proposed integrated service discovery framework adapts and makes key extensions to NMF. It enables the amalgamation of NMF with other key evidences that may be helpful to further boost the clustering accuracy. In particular, we identify two important sources of information, users’ service invocation data and extended semantics of service descriptions, and seamless integrate them into the proposed service community discovery framework. We apply the proposed framework to real-world Web services to demonstrate the effectiveness in service community discovery.