過去傳統微網誌(Microblog)的推薦方式往往忽略消費者喜好及興趣的變化速度。但消費者的喜好可能隨著時間改變而產生變化，過去相關推薦系統研究中少有研究考慮時間因素，導致推薦系統無法在適合時間點，提供適合的產品或服務給予消費者。本研究提出三種以噗文資料做的推薦方式：Personalized、Expert、Popularity，推薦契合使用者可能感興趣的產品或服務，並比較Interest-Based的Personalized、Expert及Popularity此三種推薦方式，何種方式能夠明顯提高推薦系統的推薦效果。透過本研究所提出的個人化推薦方式，相較於其他推薦方式，個人化推薦 (Personalized Recommendation)能夠有效表達使用者的個人興趣及喜好，進而提升推薦產品的點擊率，並且點擊率優於其他種推薦方式。另外本研究透過更新使用者的個人化興趣清單方式，驗證到使用者的點擊率可能隨著時間而有所改變，能夠有效找出使用者興趣變化的時間模式，以利未來行銷商隨時掌握使用者的興趣變化。Recommendation systems based on microblogs can catch users interest from message typed in social network sites.. Howerve, user preferences may change over time. To the best of our knowledge, no other work has investigated the effect of freshness on blogs when discovering user interest..The effectivess of of freshness is measured with the user clicking rate on recommendation pages of books which are related to users interest. To compare and constrat the effectinvess of interest discovering methods, this study experimented with three approaches to discover user interest, namely, Personalized、Expert and Popularity. With the Personalized approach, the key words of users in his/her microblog is identified as interested topic. With the expert approach, an expert with 10 year expereince of using microblogs was hired to analyze blogs and idenitfied interests generic to all friends and fans of a robot in Plurk. The Popularity approach is getting user interest from the top sell list of the largest internet book store in Taiwan. The approach based on personalized interest is the most effective method to catch user inerest. The expereiment also shows that user interest indeed changed with time and the words represnting personalized interest need to be revised every week to keep them effective.
