基于属性偏好和邻居信任度的协同过滤推荐算法  

Collaborative Filtering Recommendation Algorithm Based on Attribute Preference and Neighborhood Trust

作　　者：金志刚 朱琦 刘晓辉[2] Jin Zhigang;Zhu Qi;Liu Xiaohui(School of Electrical and Information Engineering,Tianjin University,Tianjin 300072,China;National Internet Emergency Center,Beijing 100029,China)

机构地区：[1]天津大学电气自动化与信息工程学院,天津300072 [2]国家计算机网络应急技术处理协调中心,北京100029

出　　处：《南开大学学报（自然科学版）》2021年第3期25-30,共6页Acta Scientiarum Naturalium Universitatis Nankaiensis

基　　金：国家自然科学基金(71502125);天津市科技计划(17KPXMSF00110)。

摘　　要：针对协同过滤算法推荐准确度低和数据稀疏的问题,提出了一种基于属性偏好和邻居信任度的协同过滤算法,首先利用用户的非共同评分项评分和项目属性信息,构建用户-属性评分矩阵,再结合共同评分项的评分计算相似度;然后利用K近邻方法获取用户的最近邻居;最后学习用户的属性偏好,结合提出的邻居信任度,计算用户的预测评分.实验结果表明,该算法有效地利用了项目属性和用户更多的评分信息,缓解了数据稀疏的问题,提高了推荐准确度.In order to solve the problems of lower recommendation accuracy and data sparsity,a collaborative filtering recommendation algorithm based on attribute preference and neighborhood trust is proposed.Firstly,the user-attribute rating matrix was calculated according to the ratings of the non-common rated items and the item attributes.And the similarity was calculated by integrating the ratings of the common rated items with the user-attribute rating matrices.Then the users’nearest neighborhoods were obtained by using the K-nearest neighbor method.Finally the user’s prediction ratings were calculated according to the user’s attribute preference and the proposed neighborhood trust.The results show that the proposed method effectively utilizes the item attributes and more rating information,alleviates the data sparsity problem and improves the recommendation accuracy.

关 键 词：推荐系统 协同过滤 项目属性 邻居信任度 稀疏问题 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]