A Hybrid Collaborative Filtering Approach for Multi-Functional Service Recommendation

As the increasing numbers of various multi-functional Cloud services are rapidly evolving in the Cloud market, how to recommend ideal multi-functional services becomes a great challenge. Due to its simplicity and promising results, collaborative filtering has been one of the most dominant methods used in recommender systems. The fundamental assumption of collaborative filtering is that users rated items similarly will rate other items similarly. However, with regards to a multi-functional service, a user may only use one of the functions and give a post-rating. Such a rating emotionally expresses the user's preference on the function rather than on the whole service. Therefore, it is appropriate to measure similarity at the granularity of functions' ratings. In our proposed approach, each user's rating is assigned to a function according to the usage record. Then, a hybrid user-based and item-based collaborative filtering algorithm is used to recommend multi-functional services. Such approach is experimentally verified at the end of this paper.