사용자 상호작용과 신뢰형성의 상호 영향에 관한 시계열 분석

초록

We are living in the era of information overload. Escaping from the information overload, online users are hoping to get reliable information. As a key solution to information overload, trust among users has received increasing attention in recent years. Trust in social media is an expression of faith and confidence that a trustor is willing to accept the information provided by his/her trustee. In social media, various applications based on trust have been developed, including services of trust-aware recommendation and finding high-quality user reviews. In reality, however, the explicit trust relations available are extremely sparse because a quite small number of users express trust relations. In order to solve the sparsity problem by inferring unknown trust relations, trust prediction methods have been proposed. Although time-stamps on trust formation and user interactions are available, no existing methods do not perform time-series analysis on them in trust prediction. In this paper, we analyze the trend of user interactions that appear before and after the trust formation on daily time-series data. We found that a certain level of user interactions is required for trust to be formed between two users. Once trust between two users has been developed, the level of user interactions between a pair of users having trust rather decreases. We expect that the result of our analysis provides a nice insight towards a more advanced approach to trust prediction.

키워드