[4H1-OS-2a] ニュースメディアのデータサイエンス(1/2)

[4H1-OS-2a-02] Adaptive RWC that is Less Sensitive to Graph Size

Yoshiki Fujikane¹, 〇Kazuhiro Kazama¹, Mitsuo Yoshida², Yoshinori Hijikata³ (1. Wakayama University, 2. University of Tsukuba, 3. Kwansei Gakuin University)

Keywords:controversy, graph, random walk

A graph-based method called RWC (Random Walk Controversy) has been proposed to quantify the controversy of social media. However, there are some problems with RWC, such as the inability to properly quantify small-scale graphs with the same sampling rate, and the lack of clear guidelines for setting the sampling rate according to the graph size. This problem becomes more serious when analyzing graphs that are mutually related but of widely different sizes, such as the relationship graphs of news media or social media audiences. In this paper, we propose a new graph-based controversy measure, Adaptive RWC, which is less sensitive to graph size and does not require user parameter setting, and show its effectiveness.