融入用户隐式信任的协同过滤推荐算法  

Collaborative Filtering Recommendation Algorithm Incorporating User Implicit Trust

作　　者：郎亚坤 王国中 LANG Ya-kun;WANG Guo-zhong(Shanghai University of Engineering Science,Department of Electronic and Electrical Engineering,Shanghai 201620,China;Key Laboratory of Artificial Intelligence Application State Administration of Radio and Television,Shanghai 201620,China)

机构地区：[1]上海工程技术大学电子电气工程学院,上海201620 [2]广播电视人工智能应用国家广播电视总局重点实验室,上海201620

出　　处：《小型微型计算机系统》2021年第8期1649-1654,共6页Journal of Chinese Computer Systems

基　　金：国家重点研发计划项目(2019YFB1802700)资助。

摘　　要：为解决传统推荐系统中的数据稀疏、关联性差、冷启动等方面的问题,有学者提出将社交中的信任关系引入推荐系统形成社会化推荐机制.这在一定程度上提高了推荐精度,但是显式信任信息很难获取并且现有的信任信息非常稀疏.针对加入用户信任信息算法的不足之处,提出了融入用户隐式信任的协同过滤推荐算法模型FITrustSVD,该模型是在TrustSVD算法的基础上融入了用户的隐式信任,定义了隐式信任用以矫正用户间的信任信息,对用户的信任范围做了约束,改进相似度算法的计算公式,并在信任预测公式中加入了用户的信任偏置.实验表明改进后的模型在数据稀疏、冷启动的条件下具有较高的推荐精度.In order to solve the problems of sparse data,poor correlation and cold start in traditional recommendation systems,some scholars proposed to introduce the trust relationship in social networks into the recommendation system to form the social recommendation.This approach improvs the accuracy of the recommendation to some extent,but the explicit trust information is difficult to obtain and the existing trust information is very sparse.Aiming at the shortcomings of the algorithm of adding user trust information,we propose a collaborative filtering recommendation model named FITrustSVD that incorporates user′s implicit trust.This model incorporates the implicit trust of users on the basis of the TrustSVD algorithm,it defines the implicit trust to correct the trust information between users and limits the range of trusted users.We improve the calculation formula of the similarity algorithm and add user's trust bias into the trust prediction formula.Experiments show that the improved model has higher recommendation accuracy under the conditions of sparse data and cold start.

关 键 词：协同过滤 社会化推荐 隐式信任 FITrustSVD算法 信任偏置 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]