Takens / / 1989/1/1 2009/9/ /1/1 2009/9/ /1/1 2009/9/30,,, i

Size: px

Start display at page:

Download "Takens / / 1989/1/1 2009/9/ /1/1 2009/9/ /1/1 2009/9/30,,, i"

さあしゃうるしはた
5 years ago
Views:

1 21 Market forecast using chaos theory

2 Takens / / 1989/1/1 2009/9/ /1/1 2009/9/ /1/1 2009/9/30,,, i

3 Abstract Market forecast using chaos theory Hiroki Hara The longitudinal data is used to forecast the market now.there is an analysis that pays attention to the analysis that uses the teacher data and the advance knowledge and the movement of data in the forecast.however, the periodicity of enough data and data is demanded from these forecasts.then, the forecast of the chaos theory is proposed.there is a Takens Embedding theorem in the forecast that uses the chaos theory.the delay value was fixation though the delay value was used in this theorem.in this research, it aims to improve the accuracy of the forecast.the delay value is made changeable for that.the chaos theory is used to forecast.it is thought that the chaos theory follows a deterministic law if the longitudinal data does chaos behavior.the chaos theory thinks that it follows a deterministic law if the longitudinal data does chaos behavior, and can forecast the behavior in the future.the longitudinal data used for the verification is collected to ranges that can be collected.the delay value and the dimension were set to the longitudinal data and it verified it.as a result, the accuracy of the forecast was able to be improved.when the dimension was changed, the difference of the accuracy of the same forecast was able to be confirmed.this research is a system that forecasts the movement of the value by using the chaos theory.it will forecast intended for the longitudinal data other than the stocks and the exchange in the future. key words Chaos theory, The longitudinal data, Time-series predicting, Changeable delay value ii

4 A 25 iii

5 / / A A.2 / A.3 / A.4 / A.5 / A.6 / iv

6 3.1 (τ = 15 ) v

7 1 [1, 2, 3, 4] [5, 6] Takens [7, 8] 1 [9] 1

8 Taakens

9 (1) (2) (3) (4) (1) 2.1 (1) (2) (1) 3

10 2.2 (3) ( ) 2.2 (a) 3 (b) 4 (c) (d) (4) 2.1 4

11 ( ) ( ) ( ) ( ) 5

12 Takens ( ) x(t) (x(t), x(t τ), x(t 2τ)),, x(t (n 1)τ)) (2.1) τ n t n n n 2 (2.2) n 2 a = (a 1, a 2,, a n ) b = (b 1, b 2,, b n ) 2.4 d(a, b) = n (a i b i ) 2 (2.2) i=1 6

16 (τ = 15 ) n / / 3 42% 52% 60% 4 62% 48% 56% 5 46% 58% 38% 6 38% 64% 42% 7 56% 46% 54% 8 60% 54% 50% 9 68% 44% 52% 3.1 τ = 15 n = 9 / n = 6 10

17 3.1 60% τ, 2τ, 3τ τ τ 1, τ 2, τ 3 (3.1) (x(t), x(t τ 1 ), x(t (τ 1 + τ 2 )),, x(t (τ 1 + τ τ i ))) t = 1, 2,, j i = 1, 2,, n 1 (3.1) 11

18 Yahoo!

19 t i = (3.2)

20 4 4.1 / / /1/1 2009/9/30 / / 1997/1/1 2009/9/ /1/1 2009/9/ / /

21 τ / / τ τ τ τ τ τ / / n = 8,

22 n / / 3 66% 64% 62% 4 62% 60% 68% 5 56% 72% 58% 6 70% 66% 60% 7 68% τ = n = 6 32 / n = 5 50% 16

23 4.3 n = 4 62% / / 17

24 / 18

25 / 19

26 5 Takens

27 21

28 1 FreeBSD

29 23

30 [1],,,,vol.40, no.8, 1997, pp [2],,,,.A,,vol.J78-A,no.12,pp [3],vol7,no.3,1995,pp [4] J.D.Farmer, J.J.Sidorowich Predicting chaotic time series, Phys.Rev.Lett., vol.59, no.8, 1987,pp [5],, Computer Today,no.99,2000,pp17-23 [6] ( ),,,,,,2000 [7] Takens,, vol.10,no.4,1998,pp [8],,,,vol.7,no.4,1997,pp [9] : B vol.129 No pp

31 A A.1 A.2 / 25

32 A.3 / 26

33 A.4 / 27

34 A.5 / 28

35 A.6 / 29

Q-Learning Support-Vector-Machine NIKKEI NET Infoseek MSN 10 1 12 22 170 121 10 9 15 12 22 85 2 85 10 i

21 Stock price forecast using text mining 1100323 2010 3 1 Q-Learning Support-Vector-Machine NIKKEI NET Infoseek MSN 10 1 12 22 170 121 10 9 15 12 22 85 2 85 10 i Abstract Stock price forecast using text