標題: 圖片導向社群網路影響力最大化之研究
Influence Maximization in Image-based Social Network
作者: 郭芷蘋
Kuo, Chih-Ping
李素瑛
Lee, Suh-Yin
資訊學院資訊學程
關鍵字: 影響力最大化;圖片導向社群網路;可擴展顏色描述子;同質性紋理描述子;Influence Maximization;Image-based Social Network;Scalable Color Descriptor;Homogeneous Texture Descriptor
公開日期: 2015
摘要: 本論文的研究目的是利用圖片的特徵建立出一個有效的圖片相似網路, 在圖片導向社群網路中解決影響力最大化問題。 本論文使用MPEG-7 標準來擷取出圖片的二種特徵值,顏色描述子使用 Scalable Color Descriptor,紋理描述子使用Homogeneous Texture Descriptor。之後再利用KMEANS 分群法計算圖片之間的相似度,建立使用者相似網路。 利用朋友資訊建立出社群網路後,再利用不同比率合併社群網路和使 用者相似網路,建構出一種新型態社群網路:圖片導向社群網路。這種圖片導向社群網路不只包含一般社群網路的社交關係,也包含圖片相似網路中圖片擁有者之間的圖片相似關係。 我們使用PGA、CDGA、CGA、DGA 四種影響力最大化演算法實驗圖片導向 社群網路。結果顯示在圖片導向社群網路中CDGA 演算法,不但能夠減少演算法的執行時間,且在提昇影響力範圍有最好的效能。
The purpose of the research in this thesis is to utilize image features to construct an effective image similarity network to tackle the influence maximization problem in image-based social network. We use MPEG-7 standard to capture two features on images. Scalable Color Descriptor and Homogeneous Texture Descriptor. KMEANS clustering method is used to calculate similarity between images for user similarity network construction. After utilizing friendship to construct social network, we merge social network and user similarity network in different ratio to construct a new type of social network – image-based social network (IBSN). An IBSN not only contains the social relations between the users in conventional social networks, but also the similarity relations between the images owned by users in the image similarity networks. We experiment influence maximization algorithms including PGA, CDGA, CGA, and DGA in IBSN. The experimental results show that the CDGA algorithm in IBSN can reduce computation time and increase influence spread. It proves that IBSN is a good solution.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT079979509
http://hdl.handle.net/11536/126957
顯示於類別:畢業論文