基于主题效能的学术文献推荐算法  

Recommendation Algorithm Based on Topic Utility for Academic Papers

作　　者：杜永萍[1] 杜晓燕[1] 姚长青[2] 

机构地区：[1]北京工业大学计算机学院,北京100124 [2]中国科学技术信息研究所,北京100038

出　　处：《北京工业大学学报》2015年第2期215-222,共8页Journal of Beijing University of Technology

基　　金：国家科技支撑计划资助项目(2013BAH21B02-01);北京市自然科学基金资助项目(4123091);ISTIC-ELSEVIER期刊评价研究中心开放基金资助项目

摘　　要：针对文献推荐问题,提出了一种基于主题效能的学术文献推荐算法,该算法使用潜在狄利克雷分布(latent Dirichlet allocation,LDA)对候选文献和用户发表的文献进行建模,挖掘出具有高效能的主题集合,并根据候选文献中高效能主题的分布情况来计算它与用户兴趣之间的相似度,最后向用户推荐有价值的文献.实验结果表明:提出的算法比基于频繁项挖掘的算法具有更高的推荐准确率和推荐召回率,可同时满足用户对个性化和文献质量两方面的需求.To solve paper recommendation problem, an academic paper recommendation algorithm based on topic utility is proposed. This approach uses latent Dirichlet allocation （ LDA ） model to build the model of candidate papers and users’ published papers, and then the topic sets with high utility are mined. The similarity between the user interest and the candidate papers is calculated according to the distribution of the high utility topics. Finally, the valuable papers are recommended to the users. Experimental results show that this method is effective, and it can get higher precision and recall than the algorithm based on apriori. Meanwhile, this method can meet the user demand for both the quality and the personalization.

关 键 词：推荐系统 效能 主题模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]