CASIA OpenIR  > 毕业生  > 硕士学位论文
基于双线性模型的标签推荐系统研究
Alternative TitleBilinear Models for Recommendation System Based on Tags
刘鹤
Subtype工程硕士
Thesis Advisor曾大军
2011-05-29
Degree Grantor中国科学院研究生院
Place of Conferral中国科学院自动化研究所
Degree Discipline软件工程
Keyword标签 双线性模型 协同过滤 推荐系统 Tags Bilinear Model Collaborative Filtering Recommendation System
Abstract随着互联网的快速发展,不断增长的信息远远超出了人们的处理能力。人们常常感觉自己淹没在信息的海洋里而无法找到自己所需要的资源,“信息过载”问题日益严重。虽然现有的搜索引擎如谷歌(Google)、百度能通过输入关键字来帮助用户过滤信息,然而这些搜索引擎返回的结果仍然是一个庞大的资源集合,同时,在更多的时刻,用户并没有明确的目标。为了给用户提供满意的信息和服务,推荐系统应运而生。其中协同过滤是最主要的推荐技术之一,其利用用户间的喜好相似性来产生推荐。 基于标签系统的网络服务和社区的发展,使得仅仅使用用户和各资源项目间的简单关系来描述的传统的推荐技术无法满足推荐在准确性,因此应该更多的考虑标签的语义和标签与用户及资源项目之间隐含的关系。在社会性标签系统中,标签作为用户对资源项目的喜好的高度化抽象描述而存在,因此,标签可用于提取用户和资源的特征。然而,把(用户、标签、资源)三维数据直接展开为三个二维的空间将会导致用户、标签及资源间隐含的信息丢失。综上所述,本文根据现有的融合标签信息的推荐技术,提出了基于双线性模型的推荐方法。在模型的训练过程中模拟用户选择某一产品的过程,从而捕捉丢失的关联信息。随后本文使用基于贝叶斯条件概率的方法验证了此模型的理论基础。 最后,本文通过实验分析来验证所提出的双线性模型标签推荐算法,并且同其他主要推荐算法进行了比较。实验结果表明,本文提出的基于双线性模型的标签推荐算法在推荐精度上有一定的提高。
Other AbstractWith the rapid development of the Internet, the growing amount of information has gone beyond the processing capacity of people. People often feel overwhelmed in the sea of information, but cannot find the resources they need. The problem of overloading information has become more and more serious. Although search engines such as Google and Baidu can help user filter useless information through key words, the results returned are still huge. Simultaneously, users do not have explicit objective at most of the time. In order to provide users with satisfactory service and information they actually need, recommendation system emerged. Collaborative filtering is one of one of the most important recommendation technical, which makes prediction with the similarity between user profiles. The development of tagging system service and communication make the recommendation technical based on the simple relation between users and items perform not precise enough. Considering the tags semantic and potential relationship between tags and users or items are necessary. In social tagging systems, tags exist as a highly abstract description of users preferences for items. Thus, tags could be used to extract users and resources features. However, if the data in the form of three-dimension (user, tag, resources) are flattening into three two-dimension spaces, the implicit information between users, tags and resources will be lost. In summary, based on the existing recommendation technologies, this paper has proposed a bilinear model based recommendations method. To capture the lost associated information, the proposed method fit the process of users selecting a product in its training phase. Then, this paper uses the method based on Bayesian conditional probability to verify the theoretical basis of this model. Finally, the paper validates the proposed bilinear model based tags recommendation algorithm with experimental analysis, and compares it with other major recommendation algorithms. The results of the experiment show that the proposed bilinear model based recommendation algorithm has certain improvements in recommendation precision.
shelfnumXWLW1683
Other Identifier200828009029055
Language中文
Document Type学位论文
Identifierhttp://ir.ia.ac.cn/handle/173211/7607
Collection毕业生_硕士学位论文
Recommended Citation
GB/T 7714
刘鹤. 基于双线性模型的标签推荐系统研究[D]. 中国科学院自动化研究所. 中国科学院研究生院,2011.
Files in This Item:
File Name/Size DocType Version Access License
CASIA_20082800902905(1409KB) 暂不开放CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[刘鹤]'s Articles
Baidu academic
Similar articles in Baidu academic
[刘鹤]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[刘鹤]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.