BIS: Bidirectional Item Similarity for Next-Item Recommendation

Exploiting temporal effect has empirically been shown to be a promising approach to improve the recommendation performance in recent years. In real-world applications, one-class data in the form of (user, item, timestamp) are usually more accessible and abundant than numerical ratings. In this paper, we focus on exploiting such one-class data in order to provide personalized next-item recommendation services. Specifically, we base our work on the framework of time-aware item-based collaborative filtering (ICF), and propose a sequence-oriented bidirectional item similarity (BIS) that is able to capture sequential patterns even from noisy data. Furthermore, we develop a compound weighting function that leverages the complementarity between the exponential weighting function and the user’s active session window. By applying the proposed weighting function and similarity measurement, we obtain a novel collaborative filtering method that achieves significantly better performance than the state-of-the-art methods in our empirical studies, showcasing its effectiveness in next-item recommendation.