期刊文献+

加入标签迁移的跨领域项目推荐算法 被引量:4

Cross-domaing Item Recommendation Algorithm Including Tag Transfer
下载PDF
导出
摘要 大多数推荐算法常采用基于迁移学习的跨领域推荐技术,借助辅助领域的丰富数据信息来解决传统单域推荐中普遍存在的数据稀疏等问题。但若迁移的知识比较单一,没有结合用户行为,则往往会在目标领域导致负迁移、推荐结果不佳等问题。因此,考虑结合其他知识来辅助完成目标领域的学习任务。利用用户异构行为改善推荐结果,正是近年来的新兴研究热点之一。在用户数据中,标签与用户的真实偏好相关,通常能够反映用户或项目的部分隐式特征。通过结合迁移学习及用户标签数据,文中提出了基于标签迁移的跨领域项目推荐算法ITTCF(Item-based Tag Transfer Collaborative Filtering)。该算法摒弃了在跨领域迁移推荐中仅对评分模式进行挖掘迁移的单一辅助方式,将用户行为反馈与数字评分相结合,融合了评分模式和标签这两种异构用户行为。在多个数据集中的实验结果均表明,ITTCF具有更好的RMSE和MAE值,较传统算法分别提升了1.61%~6.67%和1.97%~8.83%。 Most recommendation algorithms often use cross-domain recommendation technology based on transfer lear- ning and rich data in the auxiliary domain to solve the problems such as data sparse commonly existing in traditional single domain recommendation.However,if the transtered knowledge is relatively simple without combining user beha- vior ,it will lead to the problems such as negative transfer and poor recommendation results.Therefore,it is possible to combine other knowledge to assist the learning tasks in target domain.Using user heterogeneous behavior to improve recommendation results is one of the emerging research hotspots in recent years.For user data,tags are related to the real user preferences,which can reflect some implicit features of user or item.In light of this,this paper proposed a cross-domain item recommendation algorithm ITTCF(Item-based Tag Transfer Collaborative Filtering)based on tag transfer.Instead of single auxiliary moded of performing mining and migration for rating pattern in cross-domain recommendation,this method combines user behavior feedback and numeric ratings,and fuses two typical user behaviors:ra- ting patterns and tags.Experimental results on multiple datasets show that ITTCF has lower RMSE and MAE values,and its performance is 1.61% to 6.67% and 1.97% to 8.83% higher respectively than traditional algorithms.
作者 葛梦凡 刘真 王娜娜 田靖玉 GE Meng-fan;LIU Zhen;WANG Na-na;TIAN Jing-yu(School of Computer and Information Technology,Beijing Jiaotong University,Beijing 100044,China)
出处 《计算机科学》 CSCD 北大核心 2019年第10期1-6,共6页 Computer Science
基金 国家重点研发计划(2016YFB1200100) 中央高校基本科研业务费专项(2017JBM024)资助
关键词 迁移学习 跨领域推荐 标签 基于项目的协同过滤 Transfer learning Cross-domain recommendation Tag Item-based collaborative
  • 相关文献

同被引文献30

引证文献4

二级引证文献14

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部