Transactions on Machine Learning and Data Mining (ISSN: 1865-6781)

Volume 6 - Number 1 - July 2013 - Pages 3-18

The Study of the Role Analysis Method of Key Papers in the Academic Networks

Akira, Otsuki1, and Masayoshi, Kawamura2

1 Tokyo Institute of Technology, 2-12-1, Ookayama, Meguro_Ku, Tokyo 152-8550, Japan
2 The University of Tokyo, 2-11-16, Yayoi, Bunkyo_Ku, Tokyo 113-8656, Japan


In this study, we identified the problems of applying Guimera et al.ís [9] methods to a target network of articles in academic journals. Guimera et al. proposed both a clustering method and a role analysis model based on clustering. In concrete terms, they defined a Z-SCORE (Zi) and participation coefficient (Pi) as targets for metabolic networks. Although Guimera et al. methods were intended for application to metabolic networks, we believe they can be adapted to the citation networks formed by academic articles. We then proposed a new role analysis method and visualization system as a target of the academic article networks. Specifically, a unique algorithm is used to extract key articles from within clusters, after which role analysis is performed. The results are then evaluated by examining the availability of given academic articles. Finally, we performed a comparative evaluation of our method. Results showed that our method was able to show the movement of key paper innovation more clearly than Guimera et al.ís method.

PDFDownload Paper (348 KB)

Back to Table of Contents