ActiveCP: A Method for Speeding up User Preferences Acquisition in Collaborative Filtering Systems

Abstract

Recommender Systems enhance user access to relevant items _{formation, product} by using techniques, such as collaborative and content-based filtering, to select items according to the users personal preferences. Despite the success perspective, the acquisition of these preferences is usually the bottleneck for the practical use of this systems. Active learning approach could be used to minimize the number of requests for user evaluations but the available techniques cannot be applied to collaborative filtering in a straightforward manner. In this paper we propose an original active learning method, named ActiveCP, applied to KNN-based Collaborative Filtering. We explore the concepts of item’s controversy and popularity within a given community of users to select the more informative items to be evaluated by a target user. The experiments testifies that ActiveCP allows the system to learn fast about each user preference, decreasing the required number of evaluations while keeping the precision of the recommendations.