社群網路中社群關係之演變

Abstract

前人所提出的方法著重於使用單一關係來模擬訊息傳遞範圍，並且忽略社群網路上關係建立之目的以及使用者分享之意願。因此，在本篇論文中，我們提出一個考慮分享關係以及追蹤關係的方法來模擬不同時間點之資訊傳播的範圍與演變，同時，為了分析社群網路龐大的資料量，我們提出一個以MapReduce為基礎的資訊傳播分析模型。我們的方法首先以不同時間點的分享關係找出訊息傳遞的方式，接著利用追蹤關係來找出訊息真正傳遞的範圍，並觀察不同社群之資訊傳遞範圍。接著我們分析不同社群的傳遞範圍以探討社群間關係之演進，以及每個使用者在訊息傳播過程中所扮演角色之演進。實驗結果顯示不同型態的社群與使用者有不同的特徵，事件會影響訊息傳播的範圍，同時也會影響不同社群之關係以及參與傳播的使用者之角色分布。我們可根據傳遞的範圍及方式了解目標使用者的傳遞行為以及扮演的角色，分析不同時期社群的消長，協助我們在使用社群網路當作行銷工具時，能夠及時因應市場傳播趨勢來制訂最佳的競爭策略。

The previously proposed methods which only use either following or sharing relationship; however, it is not enough to realize the information propagation by just considering either one relationship. Therefore, in this thesis, we propose a method to analyze the evolution and relationships of communities, and identify users’ roles and role evolution. Moreover, we adopt the MapReduce platform to implement the proposed method since it can be easily scaled to any number of processing units. Our proposed method uses sharing relationships in different periods to compute sharing weights and construct the propagation model. Then it utilizes following relationships to find the influence range, analyze the evolution and relationships of communities, and identify users’ roles and role evolution. Thus, it can dynamically find the evolution of communities and users’ roles since the sharing relationship may change over time according to the tweet topics. The experimental results show that the communities and users have various characteristics for different types of social networks. Also, users’ roles and role distribution, and the relationships and evolution of communities are affected by important events. The proposed method may help firms formulate prompt and effective marketing and competitive strategies.