Extending user profiles in collaborative filtering algorithms to alleviate the sparsity problem

The overabundance of information and the related difficulty to discover interesting content has complicated the selection process for end-users. Recommender systems try to assist in this content-selection process by using intelligent personalisation techniques which filter the information. Most commonly-used recommendation algorithms are based on Collaborative Filtering (CF). However, present-day CF techniques are optimized for suggesting provider-generated content and partially lose their effectiveness when recommending user-generated content. Therefore, we propose an advanced CF algorithm which considers the specific characteristics of user-generated content (like the sparsity of the data matrix). To alleviate this sparsity problem, profiles are extended with probable future consumptions. These extended profiles increase the profile overlap probability, thereby increasing the number of neighbours used for calculating the recommendations. This way, the recommendations become more precise and diverse compared to traditional CF recommendations. This paper explains the proposed algorithm in detail and demonstrates the improvements on standard CF