Transcript PENDEKATAN NEURAL NETWORK UNTUK PEMODELAN TIME SERIES
Analisis Deret Waktu: Materi
minggu kesepuluh
Model ARIMA Box-Jenkins untuk data time series yang TIDAK STASIONER
Identification of NONSTATIONARY TIME SERIES
Estimation of ARIMA model Diagnostic Check of ARIMA model Forecasting
Studi Kasus : Model ARIMAX (Analisis Intervensi, Fungsi Transfer dan Neural Networks)
Example : Weekly sales of Ultra Shine toothpaste (in units of 1000 tubes) [Bowerman and O’Connell, pg. 478] t 1.
2.
3.
4.
5.
… … 26.
27.
28.
29.
30.
Y t 235.000
239.000
244.090
252.731
264.377
… … 517.237
524.349
532.104
538.097
544.948
t 31.
32.
33.
34.
35.
… … 56.
57.
58.
59.
60.
Y t 551.925
557.929
564.285
572.164
582.926
… … 805.844
815.122
822.905
930.663
839.600
t 61.
62.
63.
64.
65.
… ...
86.
87.
88.
89.
90.
Y t 846.962
853.830
860.840
871.075
877.792
… … 996.291
1003.100
1010.320
1018.420
1029.480
Example
: IDENTIFICATION step [stationarity and ACF]
ACF
Dying down extremely slowly Nonstationary time series
Example : IDENTIFICATION step … Difference [W t = Y t – Y t-1 ] Stationary time series
ACF
Y W t t AR(1) or ARI(1,1)
PACF
Dies down Cuts off after lag 1
Example
: ESTIMATION and DIAGNOSTIC CHECK step Y t = 3.0232 + 1.6591 Y t-1 – 0.6591 Y t-2 + a t Estimation and Testing parameter Diagnostic Check ( white noise residual)
Example: DIAGNOSTIC CHECK step … [ Normality test of residuals ]
Example
: FORECASTING step [MINITAB output]
Comparison
: ARIMA versus Trend Analysis ARIMA(1,1,0) MSE = 7.647
Plot comparison
: ARIMA versus Trend Analysis ARIMA(1,1,0) MSE = 7.647
Trend Analysis MSE = 598.212
Forecast comparison
Plot RESIDUAL comparison : ARIMA versus Trend Analysis ARIMA(1,1,0) MSE = 7.647
Trend Analysis MSE = 598.212
MINITAB command:
IDENTIFICATION
Step Plot Data stationarity data To make stationarity data ACF & PACF data to find tentative ARIMA model