A new collaborative filtering approach utilizing item's popularity

Collaborative filtering (CF) is one of the most successful technologies in recommender systems, and widely used in many personalized recommender areas, such as ecommerce, digital library and so on. However, most collaborative filtering algorithms suffer from data sparsity which leads to inaccuracy of recommendation. In this paper, we focus on nearest-neighbor CF algorithms and propose a new collaborative filtering approach. First, we suggest a new missing data making up strategy before user's similarity computation, which smoothes the sparsity problem. Meanwhile, the notion of item's popularity weight is defined and introduced into the computation. After then, when facing with new users, we also find a kind way to alleviate the difficulty in recommendation. The experimental results show our proposed approach outperforms the other existing collaborative filtering algorithms. It can efficiently smooth the inaccuracy caused by ratings sparsity, and can work well in generating recommendation for new users.