UTokyo Repository 東京大学

UTokyo Repository >
124 情報理工学系研究科 >
40 電子情報学専攻 >
1244025 修士論文(電子情報学専攻) >

このページ(論文)をリンクする場合は次のURLを使用してください: http://hdl.handle.net/2261/50144

タイトル: Detecting Web Spam from a Directed Graph of Web Sites
その他のタイトル: ウェブにおける有向サイトグラフからのスパム発見に関する研究
著者: Han, Bingshuang
著者(別言語): 韓, 冰霜
キーワード: Densely Connected
directed graph
Link Spam
発行日: 2007年2月2日
抄録: Link spam, which attempts to deceive link-based ranking algorithms of search engines by building densely connected structure between sites, has attracted the attention of researchers in year 2004 and 2005. It has been tightly connected with the success of commercial search engines (such as Google). In our research, we propose a technique for detecting link spam sites in the Web. Our method detects densely connected sets of sites from a directed graph of sites based on several patterns of directed connections, such as cycles and co-citations. We discuss which patterns are useful for detecting link spam, and show results of experiments on our Japanese web archive. The main contributions of this dissertation are outlined as follows: ・We propose a method for detecting the web spam structure based on several patterns of connections. ・We examined appropriate connection patterns and threshold for clustering the spam sites. ・We show the results of an extensive evaluation, based on 600 million sites and a manual examination of over 4000 sites.
内容記述: 報告番号: ; 学位授与年月日: 2007-03-22 ; 学位の種別: 修士 ; 学位の種類: 修士(情報理工学) ; 学位記番号: ; 研究科・専攻: 情報理工学系研究科電子情報学専攻
URI: http://hdl.handle.net/2261/50144
出現カテゴリ:025 修士論文
1244025 修士論文(電子情報学専攻)


ファイル 記述 サイズフォーマット
K-M221-1.pdf2.85 MBAdobe PDF見る/開く
K-M221-2.pdf7.85 MBAdobe PDF見る/開く



Valid XHTML 1.0! DSpace Software Copyright © 2002-2010  Duraspace - ご意見をお寄せください