基于用户偏好矩阵填充的改进混合推荐算法  

Improved hybrid recommendation algorithm based on user preference matrix filling

作　　者：郑小楠 谭钦红[1,2] 马浩[1,2] 刘武启 ZHENG Xiao-nan;TAN Qin-hong;MA Hao;LIU Wu-qi(School of Communication and Information Engineering,Chongqing University of Posts and Telecommunications,Chongqing 400065,China;Key Laboratory of Signal and Information Processing of Chongqing,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 [2]重庆邮电大学信号与信息处理重庆市重点实验室,重庆400065

出　　处：《计算机工程与设计》2020年第10期2784-2790,共7页Computer Engineering and Design

基　　金：国家自然科学基金项目(61671095)。

摘　　要：为进一步解决协同过滤推荐算法中存在的数据稀疏性问题,针对现有的填充算法中未充分考虑用户偏好和物品属性内在关联的问题以及相似度计算中存在的不合理之处提出一种改进算法。该算法根据评分数据分析出用户的偏好,计算用户对不同物品属性的偏好权重和评分均值,依据计算结果填补缺失项;根据目标用户改进相似度计算公式并得到基于用户偏好矩阵填充的改进混合推荐算法。实验结果表明,该算法可以解决数据稀疏问题,推荐精度均优于其它算法。To further solve the problem of data sparsity in the collaborative filtering recommendation algorithm,an improved collaborative filtering recommendation algorithm was proposed for the problems of insufficient consideration of correlation between user preferences and item attributes when filling the data and the irrationality in similarity calculation.The user’s preference for item attributes was analyzed to calculate the user’s preference weight of items and average rating score of items attributes,and the filling score based on the user’s preference weight of item attributes and average score of item attributes was calculated.According to the target user,similarity calculation formula was improved and an improved hybrid recommendation algorithm based on users’preference matrix filling was obtained.Experimental results show that the proposed algorithm can solve the data sparse problem and its recommended accuracy is better than that of other algorithms.

关 键 词：协同过滤 用户偏好 物品属性 矩阵填充 相似度计算 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]