2012 Fiscal Year Final Research Report
Model and inference of extreme synchrony in collective behavior of agents
| Project/Area Number |
23760074
|
|---|
| Research Category |
Grant-in-Aid for Young Scientists (B)
|
| Allocation Type | Multi-year Fund |
|---|
| Research Field |
Engineering fundamentals
|
|---|
| Research Institution | Kyoto University |
|---|
Principal Investigator |
SATO Aki-hiro 京都大学, 大学院・情報学研究科, 助教 (50335204)
|
|---|
| Project Period (FY) |
2011 – 2012
|
| Keywords | 経済物理学 / データ科学 / 社会経済システム / エージェント |
|---|
| Research Abstract |
This study proposes a method to quantify the structure of a bipartite graph with networkentropy from a statistical-physical point of view. The network entropy of a bipartite graph with random links is computed from numerical simulation. As an applic ation of the proposed method to analyze collective behavior, the affairs in which participants quote and trade in the foreign exchange market are quantified by using the data during the period from June 2007 to November 2012 (about 580,000,000 records with 1 -second resolution). The network entropy per node is found to correspond to the macroeconomic situation. A finite mixture of Gumbel distributions is used to fit with the empirical distribution for the minimum values of network entropy per node in each week. The mixture of Gumbel distributions with parameter estimates by segmentation procedure is verified by both Kolmogorov-Smirnov and Anderson-Dearling tests. The finite mixtures of Gumbel distributions which extrapolate the empirical probability of extreme events have an explanatory power with a statistically significant level.
|
