KAMISHIBAI KEYGRAPH: TOOL FOR VISUALIZING STRUCTURAL TRANSITIONS FOR DETECTING TRANSIENT CAUSES
Yukio Ohsawa, Takaichi Ito, MAYMI I. KAMATA · New Mathematics and Natural Computation · 2010
The causes of risks are hard to identify if events occur temporarily and disappear before the occurrence of their observable effects. In the face of this hurdle, transient causal events of significant effects are desired to be explained, for the safety of human life. In this paper, Kamishibai KeyGraph, a variation of KeyGraph developed to deal with sequential data, is presented as a tool to explain the causality involving transient causes. This method is applied here for two data sets: (1) newspaper text on social events, and (2) data on earthquakes in Japan. The performance of this method is hard to evaluate quantitatively due to the nature of transient events, so we partially evaluate the qualitative scenarios interpreted subjectively from the visualized maps.