A Weighted Similarity-boosted Collaborative Filtering Approach
Item-based collaborative filtering has been widely used in practice and is becoming the most promising approach in recommender systems. It predicts a users interest for a target item mainly based on his observed ratings. The underlying idea for item-based collaborative filtering is that the active user will more likely prefer similar items to those which he has preferred previously. Hence the choice of item-based similarity measure is critical for it. With the explosive growth in the amount of available online information, the inherent issue of rating data sparsity in user-item rating matrix is becoming a main factor impacting the accuracy of similarity computing. Aiming at the issue of sparsity, we propose a weighted similarity-boosted collaborative filtering approach employing a heuristic overlap factor to measure the degree of rating overlap. And as the experimental result shows, the proposed approach improves the quality of recommendation in contrast to the classic item-based collaborative filte ring.
recommender system item-based collaborative filtering rating data sparsity weighted similarity
Lei Ren Junzhong Gu Weiwei Xia
Dept. of Computer Science & Technology East China Normal University Shanghai Normal University Shang Dept. of Computer Science & Technology East China Normal University Shanghai, China
国际会议
海口
英文
92-96
2011-07-15（万方平台首次上网日期，不代表论文的发表时间）