Akinori Abe, Norihiro Hagita, Michiko Furutani, Yoshiyuki Furutani, and Rumiko Matsuoka [Abstract]

[Abstract]

Exceptions as Chance for Computational Chance Discovery

Akinori Abe^1,2, Norihiro Hagita^1,3, Michiko Furutani¹, Yoshiyuki Furutani¹, and Rumiko Matsuoka¹
1) Integrated Medical Science (IREIIMS), Tokyo Women's Medical University, Japan
2) ATR Knowledge Science Laboratories, Japan
3) ATR Intelligent Robotics and Communication Laboratories, Japan

In this paper, we analyze clinical data to model relationships between clinical data and health levels. During analyses of data, we discovered models which are important for determining health levels but cannot be extracted during machine learning process. We regard such models as chance and propose an interactive determination of such models. The obtained models can be referred to when standard models cannot correctly explain certain individual health levels.