Application of Beta Random Variables to Category-based Collaborative Filtering
Hadi Kalamati, Mohammad-R. Akbarzadeh-T, Sayyed-Ali Hossayni · 2018
Memory-based Collaborating Filtering (MbCF) is one of the most common techniques utilized by the recommender system. Despite its many advantages and wide applicability, this technique suffers high computational complexity, especially in the presence of scalability issues. Category-based Collaborative Filtering is an approach to MbCF for alleviating its mentioned drawbacks by categorizing the users' ratings under a few categories. In this paper, we propose a new Category-based Collaborative Filtering method with linear computational complexity by which the category information is modeled by the Beta random variable. Beta random variable requires only two parameters for intelligent data modeling of the bipolar (negative/positive) emotional patterns. Then, the parameters of the corresponding probability distribution function are estimated by a Bayesian estimator with linear computational complexity. The experiments on the MovieLens 1M dataset prove that utilizing the mentioned method and correspondingly compressing the information by the proposed probabilistic model improves the prediction accuracy and time in comparison with other state-of-the-art MbCF algorithms.