A context-aware collaborative filtering approach for service recommendation

It is a challenge to recommend Web services under multiple contexts. To address this challenge, we propose a context-aware collaborative filtering (CaCF) approach for service recommendation. Three types of contextual information, i.e. time, location and interest of user, are considered. In this approach, users' interests are extracted from service invocation records and represented as term-weight vectors. Neighbors are chosen according to the Cosine similarities of these vectors. Then, neighbors are filtered into close neighbors by location and time. At last, these close neighbors recommend service to a target user. We evaluate our method through comparing with other service recommendation approaches. The experimental results show that it achieves better precision and satisfaction rate than other two methods.