基于形式概念分析的Folksonomy用户兴趣识别研究
发布时间：2018-10-18 21:08
【摘要】：伴随着用户标签使用行为的盛行,Folksonomy的系统功能也被赋予了更高的要求。如何为用户提供满足其个性化信息需求的服务作为Folksonomy重要的研究领域之一,受到学术界的广泛关注。形式概念分析作为一种概念聚类技术,在识别数据集中概念的同时,能通过概念格形式进行可视化的呈现,为Folksonomy分析提供了形式化工具和理论支撑。基于形式概念分析的Folksonomy研究受到图书情报及计算机科学等领域部分研究者的关注。现阶段基于形式概念分析对Folksonomy用户兴趣的相关研究主要集中在从用户群的整体兴趣结构出发,识别具有共同兴趣的好友以及由标签使用频次情况计算用户对标签的兴趣度,并没有考虑用户兴趣在层级结构上的差异和关系,结合用户群标签使用的关联情况挖掘用户潜在兴趣的研究更是鲜见。基于这一现状,本研究展开了基于形式概念分析的Folksonomy用户兴趣识别一系列问题的研究。在形式概念分析理论知识的学习以及概念格结构分析的基础上,本研究提出了基于形式概念分析的Folksonomy用户兴趣识别的方法,具体包括：①基于单用户R-T概念格的分析,提出了基于单用户的兴趣度计算方法,该方法在考虑标签概念出现频次的同时,还考虑了概念所处的层级结构,从而使兴趣度计算过程更符合用户的认知；②提出基于单用户兴趣概念格结构进行兴趣概念推荐的具体策略。该策略在考虑用户对标签概念兴趣度的同时,结合了概念格具体层次结构,从而使推荐兴趣概念的内涵更加适中；③基于用户群U-T概念格的分析,提出通过用户兴趣间的关联规则来挖掘用户潜在兴趣的思路；④就概率因素导致关联规则失真问题进行探讨,提出关联规则判断依据置信度的改进方法,从而消除由于概率因素对关联规则造成的影响；⑤提出通过用户群间的关联规则识别用户的潜在兴趣并形成综合兴趣的方法。本研究通过豆瓣电影中标签、用户及资源数据的收集和标签、资源数据的预处理,采用案例分析、数理统计、问卷调查等多种方法,分别对所提出的基于单用户R-T概念格的兴趣识别方法和基于用户群U-T概念格的潜在兴趣关联性挖掘方法进行了原理分析和方法验证。实验结果表明：从总体上看,用户对通过单用户概念格识别出的推荐概念的兴趣度确实要高于其他标签概念的兴趣度;大量的用户调查显示,用户兴趣在一定程度上呈现出了如用户群U-T概念格提取的关联规则所示的兴趣关系；而修正置信度在一定程度上消除了概率因素对关联规则造成的影响,确保了兴趣概念关联规则提取的可靠性,为用户潜在兴趣的挖掘提供了支持。通过实证研究,验证了本研究所提出方法的有效性,从而为用户兴趣识别及推荐提供了新的思路。
[Abstract]:With the popularity of user label usage, the system function of Folksonomy has been given higher requirements. As one of the important research fields of Folksonomy, how to provide users with services to meet their personalized information requirements has attracted extensive attention from academic circles. Formal concept analysis, as a concept clustering technique, can identify concepts in data sets and visualize the concept lattice at the same time, which provides formal tools and theoretical support for Folksonomy analysis. The research of Folksonomy based on formal concept analysis has been concerned by some researchers in the fields of library information and computer science. At present, the research on Folksonomy user interest based on formal concept analysis is mainly focused on identifying friends with common interest from the overall interest structure of the user group and calculating the user's interest in the tag by using the frequency of the tag. It does not consider the difference and relationship of user interest in hierarchical structure, and the research of mining user's potential interest based on the association of user group tags is rare. Based on this situation, a series of problems of Folksonomy user interest recognition based on formal concept analysis are studied. Based on the theoretical knowledge of formal concept analysis and the analysis of concept lattice structure, this paper proposes a method of Folksonomy user interest recognition based on formal concept analysis, which includes: 1Analysis based on single user R-T concept lattice; An interest calculation method based on single user is proposed, which not only considers the frequency of label concept, but also considers the hierarchical structure of the concept, so that the calculation process of interest degree is more in line with the user's cognition. 2. The specific strategy of interest concept recommendation based on single user concept lattice structure is proposed. The strategy not only considers the user's interest in tag concepts, but also combines the concept lattice hierarchy, which makes the connotation of the concept of recommended interest more moderate. 3 based on the analysis of user group U-T concept lattice, The idea of mining the potential interest of users through association rules among users' interests is put forward, and the problem of distortion of association rules caused by probability factors is discussed, and an improved method for judging association rules based on confidence is put forward. In order to eliminate the influence of probability factors on association rules, a method is proposed to identify the potential interests of users and form comprehensive interests through association rules among user groups. In this study, the collection and label of label, user and resource data, pretreatment of resource data, case analysis, mathematical statistics, questionnaire survey and so on are adopted in this study. The method of interest recognition based on single user R-T concept lattice and the method of potential interest association mining based on user group U-T concept lattice are analyzed and verified. The experimental results show that, on the whole, users' interest in recommendation concepts identified by single user concept lattice is indeed higher than that of other label concepts, and a large number of user surveys show that, To some extent, user interest presents the relation of interest shown by association rules extracted by user group U-T concept lattice, and the modified confidence degree eliminates the influence of probability factors on association rules to some extent. It ensures the reliability of the extraction of interest concept association rules and supports the mining of users' potential interests. Through empirical research, the validity of the proposed method is verified, which provides a new way for user interest identification and recommendation.
【学位授予单位】：西南大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：G254
本文编号：2280369
[Abstract]:With the popularity of user label usage, the system function of Folksonomy has been given higher requirements. As one of the important research fields of Folksonomy, how to provide users with services to meet their personalized information requirements has attracted extensive attention from academic circles. Formal concept analysis, as a concept clustering technique, can identify concepts in data sets and visualize the concept lattice at the same time, which provides formal tools and theoretical support for Folksonomy analysis. The research of Folksonomy based on formal concept analysis has been concerned by some researchers in the fields of library information and computer science. At present, the research on Folksonomy user interest based on formal concept analysis is mainly focused on identifying friends with common interest from the overall interest structure of the user group and calculating the user's interest in the tag by using the frequency of the tag. It does not consider the difference and relationship of user interest in hierarchical structure, and the research of mining user's potential interest based on the association of user group tags is rare. Based on this situation, a series of problems of Folksonomy user interest recognition based on formal concept analysis are studied. Based on the theoretical knowledge of formal concept analysis and the analysis of concept lattice structure, this paper proposes a method of Folksonomy user interest recognition based on formal concept analysis, which includes: 1Analysis based on single user R-T concept lattice; An interest calculation method based on single user is proposed, which not only considers the frequency of label concept, but also considers the hierarchical structure of the concept, so that the calculation process of interest degree is more in line with the user's cognition. 2. The specific strategy of interest concept recommendation based on single user concept lattice structure is proposed. The strategy not only considers the user's interest in tag concepts, but also combines the concept lattice hierarchy, which makes the connotation of the concept of recommended interest more moderate. 3 based on the analysis of user group U-T concept lattice, The idea of mining the potential interest of users through association rules among users' interests is put forward, and the problem of distortion of association rules caused by probability factors is discussed, and an improved method for judging association rules based on confidence is put forward. In order to eliminate the influence of probability factors on association rules, a method is proposed to identify the potential interests of users and form comprehensive interests through association rules among user groups. In this study, the collection and label of label, user and resource data, pretreatment of resource data, case analysis, mathematical statistics, questionnaire survey and so on are adopted in this study. The method of interest recognition based on single user R-T concept lattice and the method of potential interest association mining based on user group U-T concept lattice are analyzed and verified. The experimental results show that, on the whole, users' interest in recommendation concepts identified by single user concept lattice is indeed higher than that of other label concepts, and a large number of user surveys show that, To some extent, user interest presents the relation of interest shown by association rules extracted by user group U-T concept lattice, and the modified confidence degree eliminates the influence of probability factors on association rules to some extent. It ensures the reliability of the extraction of interest concept association rules and supports the mining of users' potential interests. Through empirical research, the validity of the proposed method is verified, which provides a new way for user interest identification and recommendation.
【学位授予单位】：西南大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：G254
【参考文献】
相关期刊论文 前2条
1 曹高辉;焦玉英;成全;;基于凝聚式层次聚类算法的标签聚类研究[J];现代图书情报技术;2008年04期
2 杨晶;成卫青;郭常忠;;基于标准标签的用户兴趣模型研究[J];计算机技术与发展;2013年10期
，本文编号：2280369
本文链接：https://www.wllwen.com/tushudanganlunwen/2280369.html
教材专著
