基于用户特征的文献个性化推荐系统研究  

Research Literaturs Personalized Recommendation System Based on User Characteristics

作　　者：肖诗伯[1] 郭秀英[2] XIAO Shibo;GUO Xiuying(Library,North Sichuan Medical College,Nanchong,637000,China;School of Economics and Management,Southwest Petroleum University,Chengdu,610500,China)

机构地区：[1]川北医学院图书馆,南充637000 [2]西南石油大学经济管理学院,成都610500

出　　处：《网络新媒体技术》2018年第4期24-33,共10页Network New Media Technology

基　　金：川北医学院2016年校级科研发展计划项目"大数据视野下高校图书馆资源服务研究"(CBY16-B-YB02)

摘　　要：为帮助科研用户在学术数据库中更好的发现适合自身需求的文献,改善海量文献带来的"长尾效应"和"马太效应"。设计合理的文献推荐系统架构、详细的推荐引擎工作流程。以用户的科研统计学特征、行为特征、社交网络特征为研究对象,综合运用多种算法来增强和互补,并在算法中降低热门文献权重和惩罚高相关度文献。实验结果表明该推荐系统有较好准确度,能发现适合用户的文献,并能改善相关负面效应。In order to help research users to better find out their own needs for literature in academic databases,improve the ＂ long tail effect＂ and ＂ Matthew effect＂ brought by massive literature. Through reasonable design of literature recommended system architecture and recommended engine. With user＇s research statistics characteristics,behavior characteristics,social network characteristics as the research object. Using multiple algorithms to enhance and complement each other,and reduce the weight of popular literature and punish the high correlation literature in the algorithm. The experimental results show that the recommendation system has better accuracy,can find suitable literature for user,and can improve the relevant negative effects.

关 键 词：推荐系统 用户特征 协同过滤 机器学习 个性化 文献 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]