基于满意区间的协同过滤推荐算法

Satisfactory intervals similarity-based collaborative filtering recommendation algorithm

针对传统协同过滤算法中用户的个性化评价标准导致评分值不能合理地表达用户对项目的偏好程度问题,提出满意区间的概念,并设计了一种协同过滤推荐算法。该算法首先根据用户各评分值的使用概率建立其与满意区间的映射关系,然后利用满意区间的期望与标准差计算用户间的相似度,最后计算用户对项目的满意度并根据该满意度预测评分值。实验结果表明,该算法能有效地解决用户的个性化评价标准问题,提高推荐准确率。

Among traditional collaborative filtering recommendation algorithms, the difference of evaluating criteria of users caused that the user＇s ratings couldn＇t reflect the user＇s preference reasonably. In order to solve this problem, this paper proposed the concept of satisfactory intervals（ SI）, and designed a collaborative filtering algorithm. Firstly, the algorithm es- tablished the relationship between users＇ ratings and SI. Then it calculated the similarity between users through expected value and standard deviation of SI. Finally, this algorithm rated the item by its satisfaction which was calculated before. The algo- rithm solved the problem of evaluating criteria by partitioning SI, which could be more reasonable for users to express their preference. Experimental results show that this algorithm can solve the problem effectively and achieve better accuracy of re- commendation obviously.