以不同型態影響力之個人化傾向為基礎的推薦

Abstract

隨著Web 2.0的興起，人們不再侷限於面對面的互動。人們不僅可以瀏覽其他人所分享的資訊，也可以發表自己的想法與言論。然而隨著社群媒體裡的資訊數量越來越多，資訊過載的問題使人們在找自己想要或相關的資訊上是有困難的。為了解決資訊過載的問題，大部分的推薦系統利用使用者的喜好、產品的內容或是社群朋友來幫助推薦商品給使用者；然而，使用者對商品的喜好可能會受到其他不同型態的影響力影響，例如：社群朋友的影響力、興趣的影響力和熱門程度的影響力，而這三個不同型態的影響力對每個人的影響程度是不一樣的。 本研究提出一個以不同型態影響力之個人化傾向為基礎的推薦機制來推薦Flickr的照片給使用者。因為社群朋友影響力、興趣影響力和熱門程度影響力對每個使用者的影響程度是不同的，所以我們針對使用者在Flickr裡我的最愛的紀錄進行分析，以找出每個使用者對這三個不同型態的影響力的權重，並且結合影響力分數來做照片分數的預測。本研究實驗結果顯示，在社群媒體中以不同型態影響力之個人化傾向為基礎的推薦方法確實能改善推薦的準確性。

With the rapid development of web 2.0, users not only read the information shared by others but also generate content by themselves. Among the applications of web 2.0, social networking websites continue to proliferate and the volume of content keep growing, so that information overload problem makes users have difficulty in choosing useful and relevant information. To resolve such problem in social media, most of researches only utilize users’ preference, the content of items or social influence to make recommendations. However, people’s preferences towards items may be affected by three decision factors including social friends, personal interest and item popularity. Moreover, each decision factor has different impact on each user. In this work, we propose a novel recommendation method based on different aspects of influences, including social, interest and popularity, and personalized tendency towards these three decision factors to recommend photos in a photo-sharing website- Flickr. Because these influences have different degree of impact on each user, the personalized tendencies towards these three influences are regarded as personalized weights to combine the influence scores for predicting the scores of items. The experimental results show that our proposed methods can improve the prediction accuracy and the quality of recommendation.