| 摘要: | 本研究旨在以客製化為訴求，發展出一套適應網站及網站使用者行為之動態連結推薦機制，期望此推薦知識適合泛用於各式型態之網站，達到改善網站使用上的便利性，增加網站的人性化。研究中利用實際運行於網際網路中之網站及其日誌記錄，挖掘出網站中三種不同的使用者模式，設計一混合式的預測模型。並實際建置於該網站中以供使用。
在本研究中，以一個目前上線之網站，實際建置網頁動態連結推薦機制，並且分析使用者每次使用推薦連結機制後，所點選之項目回饋給推薦機制之資訊，進而使該推薦機制逐漸調整至最佳個人化。本研究在最後利用推薦機制上線後之網站日誌記錄及推薦機制之使用率、點擊項目等資訊加以分析，所推薦符合使用者預期之連結達九成以上之準確率，因此，就以便利性而言，不論是瀏覽網站效率或是瀏覽率皆能達到有效的提升。
The main objective of this research is, for the purpose of customization, to develop a dynamic linking mechanism that helps connect certain websites with their potential users’ behavior in anticipation of it being widely used in various websites to provide web users with a more convenient and user-friendly environment. There different types of user modes are explored by making use of certain websites and their log data and a mixed prediction modal is subsequently designed and installed on these sites for use.
In our research, this mechanism is set up on a web site for which the items searched by users will be analyzed and sent back as feedback in order for an adjustment of this mechanism to best meet users’ needs. According to the websites data log and the statistical analysis of the mechanism such as the utility rate, items clicked after its establishment, the accuracy of this prediction modal could reach as high as ninety percent. Furthermore, the efficiency of web browsing and browsing rate have both had a significant progress thanks to this mechanism. |
