计算机应用与软件2016，Vol.33Issue(5)：66-71,76,7.DOI:10.3969/j.issn.1000-386x.2016.05.017
利用交叉推荐模型解决用户冷启动问题
CROSS RECOMMENDATION MODEL IN SOLVING COLD-START PROBLEM
摘要
Abstract
Cold-start problem is a critical challenge for recommendation system.Traditional recommendation systems employ transfer learning techniques for this problem,i.e.to use rating/tags information in one domain to predict users and items rating in another domain. The above transfer learning model usually assumes that there aren’t the overlapping users and items between two domains.However,in many cases a system can obtain the data of same users from different domains,which differs from the above assumption.In light of such data,this paper proposes a new cold-start model for recommendation system-crossSVD&GBDT,called CSGT.It solves the cold-start challenge of user by effectively leveraging the information of overlapping users.More specifically,the proposed method extracts features from both the users and the items,and then constructs a GBDT model for training under the above assumption.Experimental data show that in Douban dataset, crossSVD&GBDT can gain the experimental result with higher performance and stronger robustness than the traditional methods.关键词
推荐系统/迁移学习/用户冷启动/交叉推荐Key words
Recommendation system/Transfer learning/User cold start/Cross recommendation分类
信息技术与安全科学引用本文复制引用
朱坤广,杨达,崔强,郝春亮..利用交叉推荐模型解决用户冷启动问题[J].计算机应用与软件,2016,33(5):66-71,76,7.基金项目
国家高技术研究发展计划项目（2012AA 011206）；中国科学院战略性科技先导专项（XDA06010600，91318301，91218302，61432001）。 （）
