- 27 2011-

Box-Jenkins

.....

) ( Box-Jenkins

ARMA . ARIMA

2015 .

125

 Box-Jenkins ....

1%2.4

:
2 ) ( .Box-Jenkins
.

: :
-1-1 :

) ( Box-Jenkins .2015

-2-1 :
:
-

1 . 2009 . 2009-2000
2 35
:
Kirchgssner G. and Wolters J. (2007) "Introduction to Modern Time Series Analysis", SPRINGERVerlag, Berlin Heidelberg, p. 3-5.

126

 - 27 2011-

-3-1 :

) ( ) ( Box-Jenkins

. SPSS .

-4-1 :
) (
.

)
( Box-Jenkins

: ) ( : Box-Jenkins

) ( Box-Jenkins Time Series :

Analysis Forecasting and Control 1976

:
-

127

 Box-Jenkins ....

: ARIMA (Autoregressive

) . integrated moving average

-

: .

: - -

)(.
-

: .

)
(.

-1-2 ) ( :Unit Root

Box-Jenkins ) ( Stochastic Process
)( .Stationary

: Xt
:3
-

< t Z , EXt2
t Z , EXt =

): t: Z (

): (

) t Z , h Z , COV ( Xt , Xt +h ) = ( h

)): (h :h (.
.

3 GOURIEROUX C. et MONFORT A., (1990) "Sries Temporelles et Modles Dynamiques " Ed.
Economica-Paris. p.152

128

 - 27 2011-

(Trend Stationary) TS Difference ) DS

.4(Stationary
-

:TS

:DS

)
( Dickey and Fuller 1979 .1981

-1-1-2 Dickey and Fuller ):5(D.F

) (D.F
) (1 :

X t = 1 X t 1 + e t
+ 1 X t 1 + e t
+ 1 X t 1 + B t + e t

)I

Xt =

) II

Xt =

) III

: X t = X t X t 1 :
: et .White Noise Process 6

4 HENIN P.Y. (1989), "Bilans et essais sur la non-Stationnarit des sries Macroconomiques" rvue
d' conomie politique n5-p. 667,668.
5 Dickey D. and Fuller W.(1979), " Distribution of the estimators for Autoregressive Time Series
With a unit Root ", Journal of the American Statistical Association, n74: pp .427-431.

129

 Box-Jenkins ....

) H 0 : 1 = 0 ( .
1

) SE ( 1

= t Dickey and Fuller .

Dickey and Fuller ) .(1

Dickey and Fuller ).(1

- 2-1-2 Dickey and Fuller ):7(A.D.F
:

B j X t j + et
B j X t j + et
B j X t j + t + e t

j =1
p

j =1

j =1

X t = 1 X t 1 +

+ 1 X t 1 +
+ 1 X t 1 +

Xt =

X t =

)I
) II
) III

: et .
( H 0 : 1 = 0 ) : .

-2-2 :ARIMA
Box Jenkins :
- 1-2-2 ):AR(p
: P ) ( Xt , t Z
:8

6 : ) (et , tZ )
( . )(GOURIEROUX et MONFORT, 1990, p.153
7 Dickey D. and Fuller W.(1981) 'The likelihood Ratio Statistics for Autoregressive Time Series
With a unit Root", Econometrica ,n49: pp .1057-1072.

130

 - 27 2011-

) =

ti

( i X

t Z, X

: i : .
: t

(1 1 B 2 B 2 .......... p B p ) X t = t
(B) Xt = t
:B .

- 2-2-2 ):MA(q

: q ) ( Xt , t Z
:9

tZ

Xt = t 1 t 1 2 t 2 ........ q t q

: i .
:

Xt = (1 1B 2 B2 ........ q Bq ) t
Xt = H ( B) t

:
- 3-2-2 ):ARMA(p,q

: ) ( Xt , t Z ) ARMA(p,q

( B) X t = H ( B) t : 10
8 Kirchgssner G. and Wolters J. (2007) "Introduction to Modern Time Series Analysis",
SPRINGER-Verlag, p.49
p.64 9
10 SHUMWAY R.H. and STOFFER D.S. (2005) "Time Series Analysis and Its Applications".
SPRINGER, p. 93.

131

 Box-Jenkins ....

p 0, q 0 ) ( B), H ( B . ) ( B), H ( B . - ) ( X t , t Z

* :ARIMA
: ) : ( Xt , t Z
d ) ARIMA(p,d,q

11

( B)(1 B) d X t = H ( B) t :
(1 B) = : . d
d

: Xt ) ARIMA(p,d,q

X t ; t 0 ).ARMA(p,q
d

-3-2 :

ARMA . Jenkins
12Box (A.C.F) )(P.A.C.F

:
* ) AR(p
.AR

& 11 PANKRATZ A. (1983) "Forecasting with Univariate Box-Jenkins Models". JOHN WILEY
SONS, p.99
12 Box, G. E. P. and Jenkins, G. M. (1976). Time Series Analysis Forecasting and Control, 2nd ed.,
Holden-Day, San Francisco.

132

 - 27 2011-

* ) MA(q
q .
* ) (A.C.F )

.ARMA(p,q

-4-2 :

ARMA

).(BENSABER A., 1989
SPSS

-5-2 :

t
.

et

)
et = X t X t

133

 Box-Jenkins ....

-1-5-2 Ljung-Box

13

H 0 : r1 (et ) = r2 (et ) = ......... = rk (et ) = 0 :

H 1 : .
: rk (et ) . k
:

) rk2 (et
k =1 N k
K

)Q = N ( N + 2

H 0 )) Q < 02.05 ( K ( p + q
-2-5-2 :
(BERA, A.K., 1981) Jarque-Bera :

T
T
1 + ( 2 3) 2
6
24

=S

32
1 = 3 :
2
4
22

= 2

S .
H 0 : et .
13 Ljung, G.M., and Box G.E.P. (1978) "on a measure of the lack of fit in time Series models".
Biometrika, n65:PP.297-303.

134

- 27 2011-

H 1 et .
) (

Kolmogorov

SPSS

: :
49

1960 2008 .
.
1960-2008 ) (1 -

1960

96239

1970

193624

1980

324895

1990

466656

2000

564708

1961

127251

1971

216447

1981

328367

1991

482022

2001

576739

1962

119851

1972

230710

1982

342209

1992

501347

2002

602160

1963

141183

1973

251246

1983

349037

1993

501654

2003

615038

1964

152952

1974

236120

1984

390881

1994

503606

2004

613600

1965

153681

1975

241640

1985

398568

1995

504374

2005

631950

1966

153609

1976

258328

1986

416665

1996

513441

2006

635195

1967

161389

1977

265460

1987

461523

1997

502114

2007

639705

1968

159589

1978

282012

1988

422812

1998

517707

2008

656477

1969

179998

1979

292347

1989

461092

1999

525595

: 2009: 1989 1976.

):(1

135

 Box-Jenkins ....

700000
600000
500000
400000
300000
200000
100000
0
2008

2005

2002

1999

1996

1993

1990

1987

1984

1981

1978

1975

1972

1969

1966

1963

1960

) (1 1960 2008

-1-3 : :
) Dickey and Fuller (

. :
) (2 Dickey-Fuller
=0.05

2.39

0.353

II

D.F.

2.39

0.729

II

A.D.F.

0.353 < 2.39 : 0.729 < 2.39 .

) (2 .14

14
.

136

- 27 2011-

50000
40000
30000
20000
10000
0
2006

2003

2000

1997

1994

1991

1988

1985

1982

1979

1976

1973

1970

1967

1964

1961

-10000
-20000
-30000
-40000
-50000

) (2

-2-3 : :
ACF ) (3

) .MA(1 PACF ) (4
) .AR(1 ): ARMA(1,1

) (3 ACF

137

 Box-Jenkins ....

) (4 PACF

-3-3
:

-1-3-3 ) AR(1 :

) ARIMA(1,1,0 . 44

1960 2003
2004 2008
.

SPSS ):(3

) (3 SPSS

Sig.

.000

7.459

.009

-2.735-

SE

Estimate

11934.170 1599.993
.143

Constant

-.391-

Lag 1

Difference

138

X
AR

X-Model_1

 - 27 2011-

) (3
. :

(1 B)(1 + 0.391B) Xt = 11934.17 + t

:

Xt = 11934.17 + 0.609 Xt 1 + 0.391Xt 2 + t

) (5 :

) (5

-2-3-3 :

:
-

) (4 Ljung-Box
)Ljung-Box Q(18
Sig.

DF

Statistics

.827

17

11.550

139

 Box-Jenkins ....

Ljung-Box

) (6 ):(7

) (6

) (7
- :
K-S -
)( ). (5

140

- 27 2011-

) (5 Kolmogorov-Smirnov
Noise residual from
X-Model_1
N

43

Normal Parametersa

173.5805

Mean

14374.55222

Std. Deviation

.075

Absolute

.075

Positive

-.066-

Negative

Most Extreme Differences

.490

Kolmogorov-Smirnov Z

.970

)Asymp. Sig. (2-tailed

a. Test distribution is Normal.

-4-3-3 ) MA(1 :
) ARIMA(0,1,1 .
SPSS ). (6
) (6 SPSS
Sig.

SE

Estimate

.000

7.483

1593.816

11926.006

Constant

Difference

.051

2.009

.151

Lag 1

.303

X-Model_2

MA

.
:

141

 Box-Jenkins ....

(1 B) Xt = 11926.01 + t 0.303 t 1
:

Xt = 11926.01 + Xt 1 + t 0.303 t 1
) (8 :

) (8

-5-3-3 :
:
- :
) (7 Ljung-Box
)Ljung-Box Q(18
Sig.

DF

Statistics

.675

17

13.888

142

 - 27 2011-

Ljung-Box
) (9 ):(10

) (9

) (10
- :
K-S-

)( ).(8

143

 Box-Jenkins ....

) (8 Kolmogorov-Smirnov
Noise residual
from X-Model_2
N

43

126.1942

Mean

14678.99773

Std. Deviation

.081

Absolute

.081

Positive

-.061-

Negative

Normal Parameters

Most Extreme Differences

.529

Kolmogorov-Smirnov Z

.943

)Asymp. Sig. (2-tailed

a. Test distribution is Normal.

-6-3-3 ) ARMA(1,1 :

) ARMA(1,1 )ARIMA(1,1,1

. SPSS ). (9
) (9 SPSS
Sig.

SE

Estimate

.000

6.977

1712.457

11947.624

.108

-1.646-

.330

.650

-.457-

.388

Constant

-.542-

Lag 1

AR

Difference

-.177-

Lag 1

MA

X-Model_1

) ARIMA(1,1,0
.15 :

15 . .

144

 - 27 2011-

(1 B)(1 + 0.542 B) X t = 11947.624 + t + 0.177 t 1

:

Xt = 11947.624 + 0.458 Xt 1 + 0.542 Xt 2 + t + 0.177 t 1

:
:
- :
) (10 Ljung-Box
)Ljung-Box Q(18
Sig.

DF

Statistics

. 825

16

10.751

Ljung-Box

)(11
):(12

) (11

145

 Box-Jenkins ....

) (12
- :
K-S -
)( ). (11
) (11 Kolmogorov-Smirnov

Noise residual
from X-Model_1
N

43

Normal Parametersa

175.9500

Mean

14336.40698

Std. Deviation

.095

Absolute

.095

Positive

-.070-

Negative

Most Extreme Differences

.622

Kolmogorov-Smirnov Z

.834

)Asymp. Sig. (2-tailed

a. Test distribution is Normal.

146

- 27 2011-

- 7-3-3 :

). (12

) (12

)ARIMA(1,1,1

)ARIMA(0,1,1

)ARIMA(1,1,0

627574.3211

625220.5637

626602.7078

613600

2004

639202.5829

637146.5697

638681.504

631950

2005

651323.4565

649072.5758

650559.0603

635195

2006

663177.0938

660998.5819

662515.3919

639705

2007

675175.7037

672924.588

674440.8869

656477

2008

MAPE = (1 / T ) ( Xt X t* / X t ) * 100

1
2

RMSE = (1 / T ) ( Xt Xt* ) 2

:
: xt .
* : xt .
). (13
) (13
)ARIMA(1,1,1

)ARIMA(0,1,1

)ARIMA(1,1,0

16784.454

14687.197

16083.862

RMSE

2.496

2.147

2.381

MAPE

147

 Box-Jenkins ....

) ARIMA(0,1,1
RMSE .MAPE ) ARIMA(1,1,0

). ARIMA(1,1,1

-8-3-3 ) ARIMA(0,1,1 :

. Line

) ( :

Yt = 664430+ 12048t

) ARIMA(0,1,1
RMSE MAPE ) (14 ).(15
) (14 ARIMA

)ARIMA(0,1,1

line

2004

613600

625220.5637

676477.6783

2005

631950

637146.5697

688525.7211

2006

635195

649072.5758

700573.764

2007

639705

660998.5819

712621.8068

2008

656477

672924.588

724669.8497

) (15
)ARIMA(1,1,0

)ARIMA(0,1,1

)ARIMA(1,1,1

line

16083.862

14687.197

16784.454

65415.470

RMSE

2.381

2.147

2.496

10.25576

MAPE

-4-3 : :
)ARIMA(0,1,1

2008

). (16

148

- 27 2011-

) (16 ) ARIMA(0,1,1 SPSS

Sig.

.000

8.140

.030

2.240

SE

Estimate

11563.188 1420.480
1
.141

.315

Constant

X-Model_2

Difference
Lag 1

MA

.
- - 2015
).(17

) (17 %95

UCL_X

Predicted_X

LCL_X

695956.2074

667326.0148

638695.8223

2009

713596.5564

678889.2031

644181.8498

2010

730321.0862

690452.3914

650583.6967

2011

746450.0851

702015.5797

657581.0743

2012

762151.7823

713578.768

665005.7537

2013

777527.5475

725141.9563

672756.3651

2014

792644.0632

736705.1446

680766.226

2015

):(13

149

 Box-Jenkins ....

) (13

:
-1

Dickey and Fuller

-2

-3

)-(

Box-

Jenkins
)-( .

) ARIMA(0,1,1 :

(1 B) Xt = 11563.19 + t 0.315 t 1

Xt = 11563.19 + Xt 1 + t 0.315 t 1

-4

-5

150

2011- - 27

:
1-

BERA, A.K. and Jarque .C.M.(1981), "An efficient large Sample test for normality
of observations and regression residuals ", Working paper in Econometrics No
40,Australion National university, Canberra.

2-

Box, G. E. P. and Jenkins, G. M. (1976). Time Series Analysis Forecasting and

Control, 2nd ed., Holden-Day, San Francisco.

3-

Dickey D. and Fuller W.(1979), " Distribution of the estimators for Autoregressive
Time Series With a unit Root ", Journal of the American Statistical Association,
n74: pp .427-431.

4-

Dickey D. and Fuller W.(1981) 'The likelihood Ratio Statistics for Autoregressive
Time Series With a unit Root", Econometrica ,n49: pp .1057-1072.

5-

FULLER A.W. (1996) "Introduction to Statistical Time Series". JOHN WILEY &
SONS, INC, New York.

6-

Kirchgssner G. and Wolters J. (2007) "Introduction to Modern Time Series

Analysis", SPRINGER-Verlag, Berlin Heidelberg.

7-

Ljung, G.M., and Box G.E.P. (1978) "on a measure of the lack of fit in time Series
models". Biometrika, n65:PP.297-303.

8-

PANKRATZ A. (1983) "Forecasting with Univariate Box-Jenkins Models". JOHN

WILEY & SONS, New York.

9-

SHUMWAY R.H. and STOFFER D.S. (2005) "Time Series Analysis and Its
Applications". SPRINGER, New York.

10- Wei, W. W. S. (1990). "Time Series Analysis Univariate and Multivariate Methods",
Addison Wesley.
11- YAFFEE R. and McGee M. (1999) "Introduction to Time Series Analysis and
Forecasting". ACADEMIC PRESS, INC, New York.

:
1-

BENSABER A. et BLEUSE-TRILLON B., (1989) "Pratique des chroniques et de le

prvision court terme". Edition MASSON-Paris.

2-

GOURIEROUX C. et MONFORT A., (1990) "Sries Temporelles et Modles

Dynamiques " Ed. Economica-Paris.

151

 Box-Jenkins ....

HENIN P.Y. (1989), "Bilans et essais sur la non-Stationnarit des sries

Macroconomiques" rvue d' conomie politique n5-pp 661-691.

3-

MARCHAL,J.L ,SEYS ,B. , et autres (2005), "Agrgation de prvisions mensuelles

de consommations medicamenteuses l'aide d'un Modle ARIMA "Rev. Statistique
Applique ,LIII (2) , PP. 5-28.

4-

MATHIS A., (1990), "Une Approche en terme de processus stochastiques vectoriels

de la dette publique Franaise" Thse de doctorat, EHESS-Paris.

5-

:
-1
2009 : 1989 1976 .

-2 ) (.
2002 .

-3 .
2005 .

-4 :
. 1992 .

.2010/7/12

152