Neighbor selection for user-based collaborative filtering using covering-based rough sets
针对传统用户协同过滤方法无法同时兼顾准确率和覆盖率的问题，提出利用覆盖粗糙集理论去除冗余用户，再从中选择最近邻，实验表明在稀疏数据集上效果更优。
Recommender systems (RSs) provide personalized information by learning user preferences. User-based collaborative filtering (UBCF) is a significant technique widely utilized in RSs. The traditional UBCF approach selects k-nearest neighbors from candidate neighbors comprised by all users; however, this approach cannot achieve good accuracy and coverage values simultaneously. We present a new approach using covering-based rough set theory to improve traditional UBCF in RSs. In this approach, we insert a user reduction procedure into the traditional UBCF approach. Covering reduction in covering-based rough sets is used to remove redundant users from all users. Then, k-nearest neighbors are selected from candidate neighbors comprised by the reduct-users. Our experimental results suggest that, for the sparse datasets that often occur in real RSs, the proposed approach outperforms than the traditional UBCF, and can provide satisfactory accuracy and coverage simultaneously.