学位論文

ミクロエージェントを用いた経済シミュレーション

Economic simulation using multiple micro agents

jpn

2013-03-24

Not Applicable (or Unknown)

2013-03-24

修士(理学)

情報科学研究科情報科学専攻; 指導教授: 藤田悟

When reproducing economic systems by multi-agents, it is very hard to collect a number of data to generate agents’ behaviors. Therefore, the goal of this paper is to construct an economic simulation that uses data from the government statistics. It is easy and costless to collect the data from the government statistics. However, raw data of the statistics should not be directly applied to the simulation, because most of the data are too generalized to assign to the agents. This paper transforms the raw statistics data to micro data to assign agents’. Additionally, this paper applies“ domain knowledge ”to the simulation in order to approach real economy. The domain knowledge is an element that features the economic model in the simulation. Using these methods, this paper conducts two kinds of simulation. One is a consumption behavior simulation. In this simulation, we used a thesaurus structure of products as domain knowledge. Another is an economic effect simulation that uses the inventory turnover period as domain knowledge. As a result, the both simulations could reproduce dynamic economic attributes.

Thesis