基于用户角色与行为的协同过滤推荐算法

Collaborative filtering algorithm based on users role and its behavior

针对传统协同过滤推荐算法中存在评分数据稀疏性问题,以稀疏的用户打分来确定用户间的相似性可能并不准确。为此,提出了以用户行为对应一定分值代替空缺评分的方法来修正用户I-U评分矩阵,并基于用户角色以权重系数K来约束最近邻的计算。实验表明,改进的算法具有更优的推荐质量。

Aiming at sparsity of score data in the traditional collaborative filtering algorithm,the similarity among users base on this sparse ratings may not be accurate.For this reason,a collaborative filtering algorithm based on fixed I-U score matrix and weighted coefficient K to constrain the nearest neighbor calculation was proposed.The fixed I-U score matrix was presented by a certain percentile of user behavior instead of vacancies scoring.The weighted coefficient K was based on user role.Experiments show that the improved algorithm has better recommendation quality.