融合多层相似度与信任机制的协同过滤算法  

Collaborative filtering algorithm based on multi-layer mixed similarity and trust mechanism

作　　者：孔麟 黄俊[1] 马浩[1] 郑小楠 KONG Lin;HUANG Jun;MA Hao;ZHENG Xiao-nan(School of Communication and Information Engineering,Chongqing University of Posts and Telecommunications,Chongqing 400065,China)

机构地区：[1]重庆邮电大学通信与信息工程学院,重庆400065

出　　处：《计算机工程与设计》2020年第12期3405-3411,共7页Computer Engineering and Design

基　　金：国家自然科学基金项目(61671095)。

摘　　要：针对传统协同过滤推荐算法推荐精度较差的问题,提出一种融合多层混合相似度与信任机制的协同过滤算法。引入模糊集隶属函数用于修正用户的评分相似度,提取用户兴趣向量计算用户对不同类型项目的偏好程度,将二者动态融合得到用户混合相似度,将用户的混合相似度与信任度进行自适应模型融合。将算法应用于MovieLens公用数据集,实验结果表明,在数据较为稀疏时,改进算法相较于改进的余弦相似度算法,准确度提升约6.3%,与部分改进算法相比,推荐精度也有一定程度的提升。Aiming at the problem of poor recommendation accuracy of traditional collaborative filtering recommendation algorithm,a collaborative filtering algorithm based on multi-layer mixed similarity and trust mechanism was proposed.The fuzzy set membership function was introduced to correct the user’s score similarity.The user interest vector was extracted to calculate the user’s preference for different types of items.The two were dynamically merged to obtain the user’s mixed similarity.The user’s mixed similarity was obtained.Adaptive model was fused with trust.The algorithm was applied to the MovieLens common dataset.Experimental results show that the accuracy of the improved algorithm is about 6.3%higher compared with the improved cosine similarity algorithm when the data is sparse.Compared with some improved algorithms,the recommended accuracy is also improved.

关 键 词：协同过滤 数据稀疏性 模糊集 兴趣相似度 信任机制 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]