摘要
现有的基于标签的协同过滤推荐方法虽然利用标签区分信息的属性,但是并不考虑标签本身的相关性,使得多数情况下信息推荐结果倾向于热门或常用标签,影响了推荐质量。针对上述问题,引入衡量标签之间的关联程度的指标——标签相关度,并基于此计算标签与信息之间对应关系的概率,从而建立一种新的标签相关度加权的协同过滤推荐算法。利用标签相关度来解决权重偏差问题,平衡热门信息和个性化信息的权重。主要方法是建立基于标签相关度特征表示的用户和信息表示,并通过特征相似性度量方法计算标签相关度加权的信息相似度,最后采用K最近方法对用户-信息偏好进行预测。实验结果表明,该方法与表现较好的LS和LW算法相比,能够在一定程度上提高推荐的精确度和召回率,更好地满足用户的实际需求。
Most of the existing collaborative filtering recommendation methods use tags to distinguish the attributes of information,but they do not con⁃sider the relevance of the tags themselves which leads to the result that in most cases,information recommendation responses tend to be popular or commonly used tags,which affects the quality of recommendation.In order to solve the above problems,a new collaborative fil⁃tering recommendation algorithm is introduced,which is based on tag correlation degree,a measure of the degree of association between tags and information,and the probability of corresponding relationship between tags and information being calculated based on this degree.The weight deviation problem is solved by using the tag correlation degree to balance the weight of popular information and personalized in⁃formation.The main method is to establish the user and information representation based on the tag correlation feature representation,and calculate the information similarity weighted by the tag correlation degree through the feature similarity measurement method,and finally use the k-nearest method to predict the user information preference.Experimental results show that this method can improve the accuracy and recall rate of recommendation to a certain extent,and better meet the actual needs of users.
作者
杨谊
张斌
和法伟
YANG Yi;ZHANG Bin;HE Fa-wei(Department of Information Technology,School of Biomedical Engineering,Southern Medical University,Guangzhou 510515)
出处
《现代计算机》
2020年第23期10-15,31,共7页
Modern Computer
基金
教育部产学协同育人项目(No.201901035041、201901104027、201901105020)
广东省本科高校创新创业教育改革研究项目(No.2018A080904)
广东省学位与研究生教育改革研究项目(No.2019JGXM22)
广东省科技计划项目(No.2017A030304009)
广东省、南方医科大学大学生创新创业训练计划项目(No.201812121066、201812121178)
南方医科大学“质量工程”建设项目:“工学融合、能力递进”医工专业人才培养模式创新实验区。
关键词
协同过滤
个性化推荐算法
标签相关度加权
信息特征计算
Collaborative Filtering
Personalized Recommendation Algorithm
Tag Relevance Weighting
Information Feature Calculation
