Interest-Driven Web Service Recommendation Based on MFI-7

Abstract

Composing and reusing services is the main advantage of Service Oriented Software Engineering (SOSE). Faced with the large amount of Web services which are available on World Wide Web, how to select and recommend suitable Web services becomes a key issue in service computing. The most popular service recommendation technique is QoS-based Collaborate Filtering (CF) with user-service QoS matrix. However, it cannot well capture the new interests of users and handle the cold start problem. In this paper we propose an interest-driven recommendation approach which leverage User Interest Profile (UIP) to represent users' interests. UIP is generated in accordance with an ISO standard named as MFI-7. The similarity between UIPs is used to produce users' nearest neighbors. The Top-K recommendations will be generated from these neighbors' mostly-used services. To show the effectiveness of our approach, a developed Web service registry and repository platform is used as a testbed to produce preliminary evaluation results.