Feature Selection Based on Audience's Behavior for Information Filtering in Online Social Networks

Abstract

Nowadays, online social networks (OSNs) become a part of human life. People use the OSNs for different purposes such as working, entertainment and so on. However, existing OSNs feed too much information into an audience's social network page without recognized the audience's situation or the audience's desire. This problem, called information overload, is a weak point of current OSNs. Our ultimate goal is to change traditional feeding information used in existing OSNs to a new type of feeding information by filtering out inconsistent information with the audience's situation. Therefore, this paper aims to prepare necessary features used in filtering information to get consistent information with the audience's situation. We also study behavior of audience in several aspects by analysis characteristic of post. The result indicates that three features are essential for the audience's decision to allow a post to be displayed on the audience's social network page, which are n1 (topic of post), n0 (audience's current situation), and n2 (type of relationship between audience and creator). Furthermore, other factors such as the number of times which a post is reviewed by the audience, age and, career have an impact on the audience's decision.