基于社交信任的概率矩阵因子分解推荐算法  

RECOMMENDATION ALGORITHM OF PROBABILITY MATRIX FACTORIZATIONBASED ON SOCIAL TRUST

作　　者：徐上上 孙福振[1] 王绍卿[1] 鹿祥志 Xu Shangshang;Sun Fuzhen;Wang Shaoqing;Lu Xiangzhi(College of Computer Science and Technology,Shandong University of Technology,Zibo 255049,Shandong,China)

机构地区：[1]山东理工大学计算机科学与技术学院,山东淄博255049

出　　处：《计算机应用与软件》2023年第11期254-258,301,共6页Computer Applications and Software

基　　金：国家自然科学基金项目(61841602);山东省自然科学基金项目(ZR2018PF005)。

摘　　要：针对评分矩阵和信任矩阵的稀疏性以及推荐精度不高等问题,提出基于社交信任的概率矩阵因子分解推荐算法PMFTrustSVD。该文采用概率矩阵分解算法对信任矩阵进行分解,分别获得用户作为信任者和被信任者的潜在社交偏好;根据用户在作为信任者或被信任者时的偏好不同,将TrustSVD算法中的无向信任矩阵修正为有向矩阵;融合两种算法来预测用户的评分矩阵。在FilmTrust数据集上实验结果表明,该算法优于现有基准算法,能有效缓解用户信任矩阵稀疏的问题并提高推荐精度。Aimed at the increasingly sparse rating matrix and trust matrix and the low accuracy of recommendation,a recommendation algorithm PMFTrustSVD of probability matrix factorization based on social trust is proposed.The probability matrix factorization was used to decompose the trust matrix,so that the potential social preferences of users as the truster and the trustee were obtained respectively.According to the different preferences of users whether users are the truster or the trustee,the undirected trust matrix in TrustSVD was modified to the directed matrix.The two models were combined to calculate the rating matrix.Experiments on the FilmTrust dataset show that the algorithm is superior to existing benchmark algorithms.It alleviates the problem of a sparse user trust matrix and improves the accuracy of recommendation.

关 键 词：推荐算法 概率矩阵分解 TrustSVD 有向信任 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]