Abstract

Sparsity is a serious problem of collaborative recommendation approach that has a considerable effect on recommendation quality. Contextual information is introduced in traditional recommendation systems besides users and items information to reduce this problem effect. Several research works proved that incorporating contextual information may increases sparse data. For this, data mining techniques are one of the most effective solutions that have been used in context-aware recommendation systems to handle sparsity problem. This paper proposes the combination of context-similarity collaborative filtering recommendation approach with data mining techniques, as solution to this problem, and develops a novel recommendation system: Rule-based Context-aware recommender system (R_CARS).The proposed approach uses a new similarity technique that aims to reduce sparse data by filtering the initial dataset that is to use only the most useful information. R_CARS is experimented on two different datasets: DePaulMovie and InCarMusic. The results of the experiment show that this combination can overcome the sparsity problem and significantly improve the performance of recommendation.

Key words: Sparsity, CARS, rule based recommendation systems, data mining, collaborative filtering, similarity.