基于链路预测的推荐方法研究  

Research on recommendation method based on link prediction

作　　者：丁沂[1] 

机构地区：[1]武汉软件工程职业学院计算机学院,湖北武汉430205

出　　处：《无线互联科技》2017年第8期99-100,共2页Wireless Internet Technology

基　　金：武汉市市属高校产学研项目;项目编号:201310

摘　　要：推荐系统在数字化环境中能够提供有价值的服务,并且在图书、电影和音乐等在线产业中取得了巨大的商业成功。大多数推荐系统采用协同过滤算法,通过分析用户和物品之间的交互行为推理用户的兴趣和偏好。协同过滤算法的推荐效果受到数据稀疏性问题的影响很大。为了解决这个问题,文章使用一种基于图的方法探索用户和物品之间的交互。文章采用二分网络链路预测的方法对用户进行物品推荐,并与协同过滤方法进行了比较,通过在豆瓣数据集上的实验结果表明,基于链路预测的方法比标准的协同过滤方法要好。The recommendation system in the digital environment can provide valuable services, and has achieved great commercialsuccess in books, movies, music and other online industries. Most recommendation system adopt collaborative filtering algorithm, whichreasons out users＇ interests and preference through the interactive behavior between users and items. The recommendation effect ofcollaborative filtering algorithm has been greatly influenced by the data sparsity problem. In order to solve this problem, this paper usesmethod on the basis of graph to explore the interaction between a user and item. This paper applies two network link prediction method torecommend items for users, and makes a comparison with the collaborative filtering method. According to the experimental results on thebean data set, it shows that the link prediction method is better than the standard collaborative filtering method

关 键 词：推荐系统 协同过滤 链路预测 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]