【摘要】：伴隨著用戶標(biāo)簽使用行為的盛行,Folksonomy的系統(tǒng)功能也被賦予了更高的要求。如何為用戶提供滿足其個(gè)性化信息需求的服務(wù)作為Folksonomy重要的研究領(lǐng)域之一,受到學(xué)術(shù)界的廣泛關(guān)注。形式概念分析作為一種概念聚類技術(shù),在識(shí)別數(shù)據(jù)集中概念的同時(shí),能通過概念格形式進(jìn)行可視化的呈現(xiàn),為Folksonomy分析提供了形式化工具和理論支撐�；谛问礁拍罘治龅腇olksonomy研究受到圖書情報(bào)及計(jì)算機(jī)科學(xué)等領(lǐng)域部分研究者的關(guān)注�，F(xiàn)階段基于形式概念分析對(duì)Folksonomy用戶興趣的相關(guān)研究主要集中在從用戶群的整體興趣結(jié)構(gòu)出發(fā),識(shí)別具有共同興趣的好友以及由標(biāo)簽使用頻次情況計(jì)算用戶對(duì)標(biāo)簽的興趣度,并沒有考慮用戶興趣在層級(jí)結(jié)構(gòu)上的差異和關(guān)系,結(jié)合用戶群標(biāo)簽使用的關(guān)聯(lián)情況挖掘用戶潛在興趣的研究更是鮮見�；谶@一現(xiàn)狀,本研究展開了基于形式概念分析的Folksonomy用戶興趣識(shí)別一系列問題的研究。在形式概念分析理論知識(shí)的學(xué)習(xí)以及概念格結(jié)構(gòu)分析的基礎(chǔ)上,本研究提出了基于形式概念分析的Folksonomy用戶興趣識(shí)別的方法,具體包括：①基于單用戶R-T概念格的分析,提出了基于單用戶的興趣度計(jì)算方法,該方法在考慮標(biāo)簽概念出現(xiàn)頻次的同時(shí),還考慮了概念所處的層級(jí)結(jié)構(gòu),從而使興趣度計(jì)算過程更符合用戶的認(rèn)知；②提出基于單用戶興趣概念格結(jié)構(gòu)進(jìn)行興趣概念推薦的具體策略。該策略在考慮用戶對(duì)標(biāo)簽概念興趣度的同時(shí),結(jié)合了概念格具體層次結(jié)構(gòu),從而使推薦興趣概念的內(nèi)涵更加適中；③基于用戶群U-T概念格的分析,提出通過用戶興趣間的關(guān)聯(lián)規(guī)則來挖掘用戶潛在興趣的思路；④就概率因素導(dǎo)致關(guān)聯(lián)規(guī)則失真問題進(jìn)行探討,提出關(guān)聯(lián)規(guī)則判斷依據(jù)置信度的改進(jìn)方法,從而消除由于概率因素對(duì)關(guān)聯(lián)規(guī)則造成的影響；⑤提出通過用戶群間的關(guān)聯(lián)規(guī)則識(shí)別用戶的潛在興趣并形成綜合興趣的方法。本研究通過豆瓣電影中標(biāo)簽、用戶及資源數(shù)據(jù)的收集和標(biāo)簽、資源數(shù)據(jù)的預(yù)處理,采用案例分析、數(shù)理統(tǒng)計(jì)、問卷調(diào)查等多種方法,分別對(duì)所提出的基于單用戶R-T概念格的興趣識(shí)別方法和基于用戶群U-T概念格的潛在興趣關(guān)聯(lián)性挖掘方法進(jìn)行了原理分析和方法驗(yàn)證。實(shí)驗(yàn)結(jié)果表明：從總體上看,用戶對(duì)通過單用戶概念格識(shí)別出的推薦概念的興趣度確實(shí)要高于其他標(biāo)簽概念的興趣度;大量的用戶調(diào)查顯示,用戶興趣在一定程度上呈現(xiàn)出了如用戶群U-T概念格提取的關(guān)聯(lián)規(guī)則所示的興趣關(guān)系；而修正置信度在一定程度上消除了概率因素對(duì)關(guān)聯(lián)規(guī)則造成的影響,確保了興趣概念關(guān)聯(lián)規(guī)則提取的可靠性,為用戶潛在興趣的挖掘提供了支持。通過實(shí)證研究,驗(yàn)證了本研究所提出方法的有效性,從而為用戶興趣識(shí)別及推薦提供了新的思路。
[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.
【學(xué)位授予單位】：西南大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2015
【分類號(hào)】：G254

【參考文獻(xiàn)】

相關(guān)期刊論文 前2條

1 曹高輝;焦玉英;成全;;基于凝聚式層次聚類算法的標(biāo)簽聚類研究[J];現(xiàn)代圖書情報(bào)技術(shù);2008年04期

2 楊晶;成衛(wèi)青;郭常忠;;基于標(biāo)準(zhǔn)標(biāo)簽的用戶興趣模型研究[J];計(jì)算機(jī)技術(shù)與發(fā)展;2013年10期

，

本文編號(hào)：2280369