由於資訊科技的發達,人們的互動除了面對面的談話,更可以利用即時通訊軟體以及社群網站進行訊息的交流,透過偌大的網際網路以及成千上萬的使用者將訊息傳播到世界各地。若能從廣大的使用者社群當中,確認出社群網路中的權威人士,也就是具有影響力之使用者,並且透過這類使用者傳播訊息,將能大幅提高訊息的傳播效率。 本研究運用Cost-Effective Lazy Forward (CELF) 演算法確認社群網路之具影響力使用者,並透過實驗確認其有效性。首先透過使用者間之好友關係建構出節點連線模型,依照使用者好友之影響力程度,轉化為影響力機率並代入CELF演算法內,透過獨立串聯模型的擴散模式,確認出影響力較大的節點,並認定為社群網路中具影響力使用者。本研究再透過Facebook建立主題性社團,並邀請對該主題有興趣之使用者加入,以此實驗驗證此演算法的有效性。本研究假設若使用者發文後所獲得的關注度越多,代表該使用者越具有影響力。經由五週實際觀察社團的發文和關注情況並且蒐集與分析數據,再以假設檢定確認演算法與實驗的相似性,結果顯示運用CELF演算法所確認的具影響力使用者,在發文所受到的關注度確實比其他使用者為高。
Now days, people communicate not only with face to face but also with instant messages and online social networks. Millions of messages spread to the world through the Internet. Therefore, if we could identify influential users and let them disseminate information through the social networks, this way will substantially increase the message spreading efficiency. This study proposed to use Cost-Effective Lazy Forward (CELF) algorithm to identify influential users in a web social network, and the effectiveness of this algorithm was evaluated by an experiment. First the node connection model was built based on the relationship between users, the influence degrees were collected from the user survey, and they were calculated as influence probability between users. Using the independent cascade model in the CELF algorithm, this method can identify influential nodes as influential users in this web social network. The next stage is to verify the proposed method and an experiment was performed. Users joined a group with common interests in the Facebook. This study assumed that the more attention a user gets when he posts a text, the more influential he is. The time period of this experiment is five weeks, and user interactions were collected and analyzed to identify influential users. The experimental result is compared to the result from the CELF algorithm with a hypothesis test. It shows that identified influential users from the algorithm and experiment are similar.