隨著社交媒體在人們生活中扮演的角色越來越重要，吸引了各個商業圈、研究圈投入的大量資金與研究。因此，該如何有效的提高社交媒體向目的客戶群分發相關的資訊及廣告已經變成了一個非常重要的問題。且由於社交媒體的普及化，導致擁有大量知識豐富的用戶參與，因此社交媒體更可以視為一項支持用戶下決策的有力參考因素。甚至於用戶會受到自己的朋友或追蹤者的影響而受到對同類型事物很大程度的喜好度影響。但大多數的社交媒體個性化推薦服務都基於單一平 台用戶建模。這可能將會遇到 數據短缺和用戶數量不足等 問題。 在本文中，我們通過彙整兩大社群平台的用戶資訊，分別為 Facebook不公開使用者資訊 與 Instagram公開熱門帳號資訊 。並建立跨平台推薦模型作為解決方案。傳統推 薦方法通常需要對推薦目標有非常多的資訊才會有較好的推薦效果，相反的，當對目 標的特質、興趣不太了解的時候，效果就會變差。而本文提出的辦法與傳統方法不一 樣的是我們可以克服對目標資訊不足時還可以 擁 有不錯的推薦效果。 本文 透過 分析兩大目標 平台的用戶行為資訊 並使用基於內容 (Content base)的方法 考量兩者 間 的特質相關性。並嚴謹的設計實驗且透過號招實際用戶幫助我們證明此方 法的有效性。;Currently, although many recommended system applications are launched, usually these recommended applications are executed on the same platform. These single-platform recommendation systems face two challenges. The first problem is the lack of data that can be referenced and used in the recommendation system. For example, for the Facebook platform, the recommended material can only come from the Facebook community itself, not from Instagram. The second problem is the problem of insufficient number of users. For example, advertisers on Instagram can only send their ads to users on Instagram, not Facebook users. In response to these two problems, this paper proposes a cross-platform recommendation system from Facebook to Instagram. This has two advantages. First, the data of the two platforms can be integrated and complement each other, thereby greatly expanding the source and richness of recommended data. Second, Instagram advertisers can not only send ads to users on the same platform, but to Facebook users with the same preferences. This can help the system expand its customer base and help better target marketing. Finally, we use a series of experiments to prove the effectiveness of the entire method. Experimental results show that this method has a good effect on the similarity analysis of Facebook users and Instagram popular accounts, and the recommendation results also highly match the user′s preferences.
