基于信息熵的协同过滤算法

针对用户评分数据的稀疏性制约着系统的推荐质量的问题,提出了一种基于信息熵的协同过滤算法。首先定义了用户信息熵以反映用户评分分布和倾向程度;然后,利用大间隔的方法计算目标用户与其他用户的间隔距离,结合目标用户的信息熵,得到目标用户的近邻选择范围;最后,同时考虑用户的信息熵和用户间的相似性大小得到目标用户的近邻集合,以降低数据稀疏性对推荐结果的影响。试验结果表明:基于信息熵的协同过滤算法能够有效地提高推荐质量。

Entropy-based collaborative filtering algorithm

In the recommender system, the recommended quality was restricted by the sparsity of user rating data. To solve this problem, a novel entropy-based collaborative filtering algorithm was proposed. First, the definition of user entropy was given to reflect the rating distribution of users and their rating tendency degree. Then, the method of large margin was introduced to calculate the margin distance, and the neighbor selection range was determined via combining both of the active users entropy and margin distance with other users. Finally, neighbors were obtained by making full of the user entropy and the similarity between users, which could degrade the influence of the sparse rating data. Experimental results on two data sets showed that the proposed algorithm could improve the recommended quality effectively.