基于用户隐式反馈和信任度的隐语义模型推荐算法

Recommendation Algorithm of Latent Factor Model Based on User Implicit Feedback and Trust

针对传统隐语义模型(Latent Factor Model,LFM)算法只是单一使用用户对物品的评分矩阵信息来进行预测评分,对于用户和项目之间存在隐含信息并未利用,导致个性化推荐精确度很低。为了处理这个问题,设计结合用户隐式反馈和信任关系的推荐算法BI-Trust-LFM,利用用户行为中包含的隐式信息来替代用户属性,但是用户行为信息存在稀疏性,所以通过改进信任关系来更进一步进行个性化推荐。实验表明,该算法可以提高推荐系统的预测评分准确度,同时改善个性化推荐精准度。

For the traditional Latent Factor Model(LFM)algorithm,it only uses the scoring matrix information of users to predict and score items,and does not use the implicit information between users and items,resulting in low accuracy of personalized recommendation.In order to deal with this problem,a recommendation algorithm BI-Trust-LFM,which combines implicit feedback and trust relationship of users,is designed.It uses implicit information contained in user behavior to replace user attributes.However,user behavior information is sparse,so we further personalized recommendation through trust relationship between users.The experiment shows that the algorithm can improve the accuracy of personalized recommendation to a certain extent by comparing the recommendation results of other algorithms through simulation in Movielens data set.