摘要
传统协同过滤算法在利用评分进行推荐时,面临物品冷启动及不能保证评分客观公正的问题,从而影响推荐的准确度。同时,推荐系统所需数据集中存在大量用户行为信息以及物品标签信息。为了解决以上问题,综合标签信息来提出一种结合用户行为和物品标签的协同过滤推荐算法。实现了不依赖用户对物品的评分,而基于物品标签概率进行有效推荐的研究目标。实验表明,该算法可以消除不客观公正评分的影响并能很好地解决物品冷启动问题,同时提高推荐准确度。
The traditional collaborative filtering algorithm is facing the problem of cold start of items and the failure to guarantee the objective and fair scores when using the scores for recommendation,thus affecting the accuracy of recommendation. At the same time,there is a large amount of user behavior information and item labels information in the recommended data set. In order to solve the above problems,a collaborative filtering recommendation algorithm combining user behavior and item labels was proposed. It realized the research goal of effective recommendation based on the probability of item labels without relying on user ＇s rating of the item. Experiments showed that the algorithm eliminated the impact of not objective and fair score and solved the problem of cold start of items well and improved the accuracy of recommendation.
作者
李龙生
艾均
苏湛
李妍妍
Li Longsheng;Ai Jun;Su Zhan;Li Yanyan(School of Optical-Electrical and Computer Engineering, University of Shanghai jbr Science and Technology, Shanghai 200093, China;School of Electronic Communication Engineering, Zhengzhou University of Aeronautics, Zhengzhou 450003, Henan, China)
出处
《计算机应用与软件》
北大核心
2018年第6期248-253,共6页
Computer Applications and Software
基金
上海市自然科学基金项目(14ZR1428800)
关键词
协同过滤
冷启动
用户行为
物品标签
Collaborative filtering Cold start
User behavior Item labels
作者简介
李龙生，硕士生，主研领域：推荐系统，复杂网络。;艾均，讲师。;苏湛，讲师。;李妍妍，本科生。
