FSBPR: a novel approach to improving BPR for recommendation with the fusion of similarity

As an essential part of big data, the recommendation system is widely used because of its practicability. Most of the traditional rating prediction algorithms mainly focus on explicit feedback, but this type of data is usually sparse in the real world. By contrast, the Bayesian Personalized Ranking (BPR) algorithm could directly optimize for ranking and provide personalized recommendation from implicit feedback. The BPR is a well-known pairwise method for one-class collaborative filtering, it proposed two assumptions: (1) it assumes all the unrated items are negative items for each user; and (2) it assumes user prefer rated items to unrated items. However, the assumptions in BPR may not always hold in reality. That is because for the unrated items, a user may have different view, such as potentially like or dislike. To mitigate the above-mentioned problems, we propose a novel approach to improving BPR for recommendation with the fusion of similarity, termed as FSBPR for brevity. We first fuse the predefined similarity and the learned similarity method to calculate the similarity between items, and then find the items that users may be interested in through the number of item occurrence. To this end, we divide the items into three sets for each user and provide the pairwise preference assumption. We also extend the idea of the CoFiSet model to our model and provide a series pairwise assumption. The model parameters are learned by the stochastic gradient descent. We conduct experiments on three real-world datasets to verify the accuracy of FSBPR, and compare FSBPR with the state-of-the-art methods. The experimental results indicate that our model significantly improves the accuracy of the recommendation.