19^th International CODATA Conference
Category: Poster, Knowledge Discovery

An Approach for Data - Driven Decision Making

Masahumi Yabe (yabe@qs.t.u-tokyo.ac.jp), Nobuto Oka, Yasuyuki Nakai,and Shuichi Iwata
The University of Tokyo, Japan

Choosing an appropriate answer in a finite time is a problem we always face every field. For that, it is most important to capture data of high quality as much as possible, which can be basic for better decision making. But we can not collect perfect series of data in the real world, because of the limitation of time, cost, and size of data, and decision making must be done under the condition that only insufficient information are gotten, and proactive converging methods for searching more appropriate solution for it is requested.   

In case of the system that is irreversible as environmental problem or many trials can not be allowed because of cost problem, this proactive approach becomes especially valuable. There is a possibility that just one action may change all of situations and the semantics of data which were collected before the fact happens. So we always have to analyze how strong a selected action makes influences to the system before making the last decision.

In this research, we propose a data-driven method for exploring an effective action under the condition that data are spatially dispersed, namely, a collection of dots. A database on materials is used as a starting point for this research as its data model has a hierarchy and plural facets on one fact.