2012 Fiscal Year Final Research Report
Studying the methodology of analyzing non-stationary time series by taking high-frequency price data as an example
| Project/Area Number |
20510137
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Social systems engineering/Safety system
|
|---|
| Research Institution | Tottori University |
|---|
Principal Investigator |
TANAKA Mieko 鳥取大学, 大学院・工学研究科, 教授 (20257570)
|
|---|
| Project Period (FY) |
2008 – 2012
|
| Keywords | 非定常時系列 / 経済物理 / 主成分分析 / ランダム行列 / 乱数度評価 / RMT-PCA |
|---|
| Research Abstract |
We examined the methods of analyzing non-stationary time series by using high frequency price time series. In order to treat the strong randomness and the time dependence, we avoided taking averages of the data but treated the original time series directly by solving the eigenvalue problem of the cross correlation matrix between pairs of price time series to compare its eigenvalue distribution to the corresponding RMT formula. In this study, we have successfully established the application of "RMT-PCA" on stock markets, and also invented a new tool to measure the randomness, that we call the "RMT-test", and proved its effectiveness by comparing the levels of randomness of various random generators, measuring the quality of hash functions, and also the choice of stocks to invest.
|
