前端

主要使用ASP、Html、JQuery、JavaScrip、CSS為主。曾應用的套件有：Vue.js、Ext.js、D3.js、JStree、Kendo、FancyBOX、Select2、SweetAlert、BootStrap、DataTable

後端

主要使用c#、Python、PHP、RDL(SSRS)。

熟悉MVC、Entity Framework、LINQ、Data Mining、WCF。

曾使用過的版控軟體有：SVN、Git

資料庫

曾使用MSSQL做過大型專案開發(ERP)、實作MySQL於專題及論文

雲服務

AWS S3、AWS CloudFront、AWS EC2、Vimeo影音串流、第三方金流

在現今人們越來越依賴網路的時代，且網路上時刻存在著海量的資訊，人們需要一個能便於閱讀、儲存及篩選文章的系統，因此，社交書籤網站應運而生，社交書籤網站蘊含了許多人們豐富的知識產物及資訊片段。 由於每個人的興趣皆有所不同，所關注的議題也不一樣，在這個資訊爆炸、時間即時金錢的時代，從複雜且海量的資訊當中如何過濾出對自己有用的資訊成為新一代的課題。因此，本研究所探討的中點即是，使用者該如何從自己的社交圈中快速且有效的找出與自已志趣相投的物件，又如何多元的接收新知。節省大量的人力與時間，做最有效率的學習。 本研究基於上述想法，為改善以知識為基礎的人際社交網路，利用計算兩個使用者之間的相似度，探討各人際關係層間與目標使用者的相似度，找出與目標使用者興趣相似的使用者。再藉由目標使用者社群關係，推薦目標使用者可能感興趣的文章。 本研究藉由計算餘弦相似度驗證人與人、文章與文章、人與文章之間的關聯性。本研究所提之推薦方法準確度上優於隨機推薦模式、依時間順序推薦模式，多樣性及執行時間上優於依相似度推薦模式。 本研究所提之推薦方法有效的過濾了大量使用者可能不感興趣的資訊，並在短時間內提供與使用者相關且多樣的文章資訊，提供使用者閱讀。使用者也能快速的了解自己所關注的人，最近在關注什麼事物，以及是否有相似度較高得追隨者(follower) 值得關注(follow)，使相似度較高的追隨者與目標使用者的關係變成雙向追隨，進而提高文章推薦的優先權。 

關鍵詞：社交網路、推薦系統、Tagging、Diigo

Plenty of information appears on the Internet everywhere all the time. People need a system to read conveniently, store easily, and select fast because of the enormous amount of information. Therefore, the social bookmarking built, containing plentiful knowledge. In today’s society, how to choose useful information is a significant issue for the new generation. In fact, the hobbies are different from people to people and the subject people care will also be different. As a result, the main point of this research is that how users find out what they are interested in and how they receive new knowledge from their society circles quickly and efficiently with low manpower and less time. To improve the knowledge-based interpersonal society networks, this study uses cosine similarity to calculate the similarity between relationships and target users in the research and then finds the users who the target users are interested in. Eventually, it can recommend the articles to the target readers who may be interested in it by their community relationships. Calculating cosine similarity proves the cognation between people, between articles, and between people and articles. The model in the research is more accurate than the model that recommend randomly and chronologically. Furthermore, the diversity and executing time of the model in this study are better than the model of cosine similarity. The model which this research recommends filters a great amount of useless information efficiently and provides users with relative and diverse information in short time. Users can also know what their followings care about and whether there are followers with more similarity to follow. Then, the relationship between target user and followers with more similarity becomes two-way and it promotes the recommending priority of articles. 

Key word：Social network、Recommend system、Tagging、Diigo