本文選題：鏈接推薦　切入點(diǎn)：社區(qū)發(fā)現(xiàn)　出處：《哈爾濱工業(yè)大學(xué)》2014年碩士論文　論文類型：學(xué)位論文

【摘要】：以往關(guān)于鏈接推薦方面的研究，大部分關(guān)注于對(duì)社會(huì)交往功能的加強(qiáng)，而忽視了對(duì)信息擴(kuò)散功能的加強(qiáng)。鏈接推薦算法不應(yīng)該僅僅專注于評(píng)估用戶信息間的相似度或者社交關(guān)系間的相似度，同時(shí)也應(yīng)該關(guān)注那些應(yīng)用了推薦算法在網(wǎng)絡(luò)中增加的邊，使得信息在更新的網(wǎng)絡(luò)結(jié)構(gòu)上獲得擴(kuò)散的最大化。前人通過引入社區(qū)發(fā)現(xiàn)提出了結(jié)點(diǎn)擴(kuò)散度的概念，，并將結(jié)點(diǎn)擴(kuò)散度的結(jié)果和傳統(tǒng)鏈接預(yù)測(cè)結(jié)果相結(jié)合來解決上述問題。 本文的研究工作主要分為以下三方面： 1.首先給出了鏈接推薦用于增強(qiáng)信息擴(kuò)散的研究框架，并選定了框架中核心部分相對(duì)應(yīng)的算法和模型，對(duì)比分析了現(xiàn)有的三種計(jì)算結(jié)點(diǎn)擴(kuò)散度的方法； 2.分析現(xiàn)有結(jié)點(diǎn)擴(kuò)散度算法的不足，通過引入社區(qū)對(duì)結(jié)構(gòu)提出了非重疊社區(qū)結(jié)點(diǎn)擴(kuò)散度改進(jìn)版算法； 3.分析重疊社區(qū)的信息擴(kuò)散結(jié)構(gòu)，通過引入結(jié)點(diǎn)的社區(qū)中心性提出了重疊社區(qū)的結(jié)點(diǎn)擴(kuò)散度算法。 本文在真實(shí)的無向社交網(wǎng)絡(luò)數(shù)據(jù)集Email-Enron及Amazon和有向社交網(wǎng)絡(luò)數(shù)據(jù)集Email-EuAll上進(jìn)行了對(duì)比實(shí)驗(yàn)，結(jié)果在緊密型網(wǎng)絡(luò)上超過了現(xiàn)有方法對(duì)信息擴(kuò)散的促進(jìn)。 本文更加注重結(jié)點(diǎn)的社區(qū)屬性對(duì)信息擴(kuò)散的影響，完善了結(jié)點(diǎn)擴(kuò)散度解決方案體系，大幅度降低了計(jì)算復(fù)雜度，使得可以適用于大規(guī)模社交網(wǎng)絡(luò)，另外本文還通過實(shí)驗(yàn)引出了兩個(gè)值得關(guān)注的問題，有助于后人對(duì)結(jié)點(diǎn)擴(kuò)散度方法體系的進(jìn)一步研究。
[Abstract]:Most of the previous studies on link recommendations have focused on strengthening social interaction, Link recommendation algorithms should not only focus on evaluating the similarity between users' information or social relationships, but also on the edges that are added to the network by using the recommendation algorithm. By introducing community discovery, the concept of node diffusivity is put forward, and the results of node diffusion are combined with the traditional link prediction results to solve the above problems. The research work of this paper is divided into the following three aspects:. 1. Firstly, the research framework of link recommendation to enhance information diffusion is given, and the corresponding algorithms and models of the core parts of the framework are selected, and three existing methods to calculate the diffusion of nodes are compared and analyzed. 2. Analyzing the deficiency of the existing node diffusivity algorithm, an improved algorithm is proposed by introducing community to the structure of non-overlapping community node diffusivity. 3. The information diffusion structure of overlapping communities is analyzed, and the algorithm of node diffusion degree is proposed by introducing the community centrality of nodes. This paper makes a comparative experiment on the real undirected social network data set Email-Enron and Amazon and the directed social network data set Email-EuAll. The results show that the information diffusion on the compact network is more than that of the existing methods. This paper pays more attention to the effect of the community attribute of the node on the information diffusion, improves the solution system of node diffusion, greatly reduces the computational complexity, and makes it suitable for large-scale social networks. In addition, this paper raises two problems worth paying attention to through experiments, which is helpful to the further study of the method system of node diffusivity.
【學(xué)位授予單位】：哈爾濱工業(yè)大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2014
【分類號(hào)】：TP393.0

【參考文獻(xiàn)】

中國(guó)期刊全文數(shù)據(jù)庫(kù) 前3條

1 駱志剛;丁凡;蔣曉舟;石金龍;;復(fù)雜網(wǎng)絡(luò)社團(tuán)發(fā)現(xiàn)算法研究新進(jìn)展[J];國(guó)防科技大學(xué)學(xué)報(bào);2011年01期

2 許曉東;肖銀濤;朱士瑞;;微博社區(qū)的謠言傳播仿真研究[J];計(jì)算機(jī)工程;2011年10期

3 李棟;徐志明;李生;劉挺;王秀文;;在線社會(huì)網(wǎng)絡(luò)中信息擴(kuò)散[J];計(jì)算機(jī)學(xué)報(bào);2014年01期

本文編號(hào)：1562610