社群網路上具資訊價值之朋友推薦方法

Abstract

社群網站使用者們利用自己的社交動態更新分享資訊，而朋友們的動態則形成能讓使用者接收到最即時資訊的社交動態流(social update stream)，然而，在社群網站擁有過多的朋友會導致資訊過載(information overload)的問題。 本研究提出了一個朋友推薦模型，利用矩陣分解和排序學習法以模擬使用者和動態訊息各自的潛在偏好，同時，也考慮社群影響力此一因子以強化潛在偏好的學習能力，本模型將會依照目標使用者之偏好與所有候選人分享的動態訊息集合之潛在偏好契合程度推薦出有價值的朋友。本研究於現實世界的巨量資料集進行實驗，所建立之朋友推薦模型在覆蓋率和排序表現上都大大超越了常見的朋友推薦方法，實驗結果展現社群影響力和排序學習法在社群網站上推薦朋友之有效性。

Social network sites have gradually taken the place of traditional medium for people to receive the latest information. To receive novel information, users of social network sites are encouraged to establish social relations. The updates shared by friends form social update streams that provide people with up-to-date information. However, having too many friends can lead to an information overload problem causing users to be overwhelmed by the huge number of updates shared continuously by numerous friends. The information overload problem may affect user intentions to join social network sites and thereby possibly reduce the sites’ advertising earnings which are based on the number of users. In this paper, we propose a learning-based recommendation method which suggests informative friends to users. A user is considered as an informative friend if people like the updates the user posts. Techniques of learning to rank are designed to analyze user behaviour and to model the latent preferences of users and updates. At the same time, the learning model is incorporated with social influence to enhance the learned preferences. Informative friends are recommended if the preferences of the updates that they share are highly associated with the preferences of a target user. The experiment results based on a huge real-world dataset demonstrate the effectiveness of combining learning to rank with social influence on the informative friend recommendation task. The proposed method is effective and it outperforms many well-known friend recommendation methods in terms of the coverage rate, mean reciprocal rank, and the discounted cumulative ranking performance.