高校图书馆个性化图书推荐系统应用研究-中国科技论文在线.PDFVIP

第8 卷第8 期 Vol.8 No.8 2015 年4 月 April 2015 高校图书馆个性化图书推荐系统应用研究 * 于海涛, 闫相斌 （哈尔滨工业大学图书馆，哈尔滨 150001 ） 摘要：针对目前个性化推荐系统中仍然存在的关系建模、数据稀疏等问题开展研究。提出面向高校图书馆的个 性化图书推荐系统框架，对用户之间、图书之间的同质性关系，用户和图书之间的异质关系进行建模，尽可 能解决数据的稀疏性问题。推荐系统采用正交非负矩阵三分解实现联合聚类算法，确保解的唯一性和可解释 性。联合聚类算法同时对用户和图书进行聚类分析，进而提高聚类的准确性。通过历史数据的对比分析发现， 提出的推荐算法在准确性方面有了较为明显的提高。 关键词：图书馆学；数字图书馆；个性化推荐；联合聚类；数据稀疏 中图分类号：G250.7 文献标识码：A 文章编号：1674-2850(2015)08-0758-05 Application research on personalized book recommendation system of university library YU Haitao, YAN Xiangbin (Harbin Institute of Technology Library, Harbin 150001, China) Abstract: Aiming at the personalized recommendation system still existing in the relationship modeling, data sparse research and other issues, we proposed the university library personalized recommendation system framework for users, books homogeneity relationship between the user and book modeling, the relationship between heterogeneity, possibly to solve the problem of data sparsity. Recommended system uses three orthogonal non-negative matrix decomposition algorithm for joint clustering to ensure the uniqueness of solution and interpretability. Joint clustering algorithm analyzes books and users simultaneously to improve the accuracy of clustering. Through comparative analysis of historical data, The accuracy of the proposed recommendation algorithm is significantly increased. Key words: library science; digital library; personalized