Title: 分析臉書粉絲團資訊以發展使用者特徵檔為基礎之推薦
Recommendations based on user profiles discovered from Facebook Like List
Authors: 李榮維
Lee, Rong-Wei
劉敦仁、林君信
管理科學系所
Keywords: 社交網路;推薦系統;正規化概念分析;Social Network;Recommendation System;Formal Concept Analysis
Issue Date: 2012
Abstract: 近年來社群網路的蓬勃發展,除了傳統的部落格網站,許多新型態的社群網站因應而生,如微型網誌Twitter、Plurk、社交網站Facebook、MySpace等,隨著這些社群網站的興起,大大改變了人們原本的使用網路的習慣。 社群網站-Facebook在使用的過程中,如張貼訊息、回應訊息、對訊息點選「讚」、訂閱粉絲團等資訊,都間接透露出使用者的需求與喜好等資訊。 本研究欲利用使用者訂閱的粉絲團清單,找出使用者的興趣關鍵字特徵檔,並進行相關應用:推薦粉絲團、推薦朋友、推薦廣告。本研究首先利用「正規化概念分析法」分析使用者訂閱的粉絲團之間彼此的關係,進而發掘代表使用者興趣的概念,產生使用者興趣關鍵字特徵檔,最後利用產生的興趣關鍵字特徵檔進行推薦。 本研究實地蒐集了Facebook的使用者資料以及Yahoo奇摩拍賣的內容式廣告建立雛形系統,以雛形系統展示研究成果及模擬使用者操作推薦粉絲團、推薦朋友之情況和推薦廣告的成果。
After the fever of web 2.0 when blogging embarked on a new era of internet community, social networks have flourished in recent years and become another Internet golden era. Besides the traditional blog website, many new types of social network websites such as Facebook, MySpace or Microblog Twitter, Plurk came up. Social network websites have changed user behaviors on Internet nowadays dramatically. In facebook, personal interests or needs are disclosed when users post messages on the wall, reply friends’ messages, push “like” messages or join “fan pages”. This research proposes to recommend fan pages, friends and advertisements through discovering user interests from Facebook users’ “like list”. In this thesis, Formal Concept Analysis (FCA) is adopted to analyze the relation between fan pages in the like list. Afterwards, concepts that are appropriate to represent users’ interests are extracted to generate user profiles. Finally, a prototype System is developed to demonstrate the research result of recommending friends, fan pages and advertisements based on the discovered user profiles.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT079831535
http://hdl.handle.net/11536/47806
Appears in Collections:Thesis


Files in This Item:

  1. 153501.pdf

If it is a zip file, please download the file and unzip it, then open index.html in a browser to view the full text content.