標題: 社群網路資料流動分析模型之研究:以Facebook為例
Information Diffusion Model of Online Social Networks: A Case Study of Facebook
作者: 陳玟媗
CHEN, WEN-HSUAN
黃育綸
Huang, Yu-Lun
電控工程研究所
關鍵字: 社群網站;Facebook
公開日期: 2013
摘要: 社群網站提供使用捨一個分享及取得資訊的新平台,使用者之間的互聯杏因其互動行為變動而展現不同的強度關係。現有研究主要探討如何透過使用者在虛擬世界的互動關係,及個人資訊,逆推得到真實環境之人際關係。此外,也有其他研究嘗試探討社群網路在資訊擴散中所扮演的角色。然而,這些研究並未討論虛擬人際關係對資訊流動的影響,例如當某個使用者被植入殭屍病毒,則該使用者有可能會變成攻擊者,散播惡意連結給他的朋友,並進而造成大範圍的網路感染。在此論文當中,我們針對資料流動提出一個分析方法。我們的方法包含兩個階段:第一階段我們先建立分析模型,利用使用者間的互動,計算出兩兩使用者的關係強度與回應比例。透過第一階段所得到的關係強度與回應比例,我們在第二階段進一步地建立單點及多點資訊擴散模型,並預測資料的流動路徑、評估該使用者對資訊擴散的影響。以Facebook為研究案例,我們利用Facebook上實際的數據,以驗證我們所提方法的正確性。實驗結果顯示,我們的方法可以預測每個受測者所張貼的資訊的可能擴散範圍。相較於現有的研究,我們的方法可以找出影響資料擴散的弱連結,並篩選出對資訊擴散影響力最大的使用者群。
Online social networks (OSNs), like Facebook, Twitter and Google+, have created a novel way for people to connect with each other by sharing and obtaining information. As user behaviours vary, their inter-connectivities also reveal different levels of tie strengths within the OSNs accordingly. By researching the similarities of user profiles and interactions, some existing works have presented the formal analysis to explain dyad relationships; some others have explained OSNs roles in information diffusion. However, none of them addresses the security impact resulted from a user who tries to spread his information. If such a user is attacked with his post injected by any malicious link, the user then becomes a bot master to disperse the information to his friends and potentially taint the network at large scale. In this thesis, we propose a method to predict the information diffusion of a post within an OSN. In addition to tie strengths of dyads, we claim that responding patterns and rates, along with the involvements of friend’s friends, should also be taken in account when predicting the possible delivering paths of information. In our method, we first estimate the one-hop information diffusion, then we can derive multi-hop information diffusion accordingly. We conduct several experiments to verify the validity of our predictions with the real data obtained from the Facebook website. The experiment results show that our method predicts the diffusion coverage of a piece of information from a ii specified individual.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT070060002
http://hdl.handle.net/11536/74109
Appears in Collections:Thesis