frequency domain techniques i n forecasting with arima model
DESCRIPTION
Frequency Domain Techniques i n Forecasting with ARIMA Model. Dr. M. J ahanur Rahman Associate Professor Department of Statistics University of Rajshahi , Rajshahi-6205, Bangladesh [email protected] July 28, 2012. Contents. Time Series? Time domain forecasting with ARIMA model - PowerPoint PPT PresentationTRANSCRIPT
![Page 1: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/1.jpg)
Frequency Domain Techniques in
Forecasting with ARIMA Model
Dr. M. Jahanur RahmanAssociate Professor
Department of StatisticsUniversity of Rajshahi, Rajshahi-6205, Bangladesh
July 28, 2012
![Page 2: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/2.jpg)
Dr. M. Jahanur Rahman, R.U.
Contents
Time Series?
Time domain forecasting with ARIMA model
Transformations from time domain to
frequency domain
Frequency domain techniques in forecasting
with ARIMA model
2
![Page 3: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/3.jpg)
Dr. M. Jahanur Rahman, R.U. 3
Time Series
A sequence of observations indexed by time, denoted
by or
Monthly relative humidity time series of Rajshahi
![Page 4: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/4.jpg)
Dr. M. Jahanur Rahman, R.U. 4
π¦ π‘+2
π¦ π‘+1π¦ π‘
π¦ π‘β1
π¦ π‘β2
= Realization
orSample pathTime Series?
A finite realization of a time series process
Let be a sequence of random variable indexed by time t.
Random Process to Time Series
![Page 5: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/5.jpg)
Dr. M. Jahanur Rahman, R.U. 5
Note that
Time series is only one of the many possible
realizations (sample paths) that the history might
have generated.
Time series analysis is trying to draw statistical
inference from a single outcome (time series),
even partially observed.
![Page 6: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/6.jpg)
Dr. M. Jahanur Rahman, R.U. 6
How can we make sensible
inferences on the underlying
time series process with a single
observation?We need a strong assumption:
Stationarity (or ergodicity). This assumption treat the time series as a random sample
from the same underlying population (process).
![Page 7: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/7.jpg)
Dr. M. Jahanur Rahman, R.U. 7
A time series is weak stationary if its mean, variance, and autocovariance (at various lags) remain the same no matter at what point of time we measure them. That is, they are time invariant.
Stationary Time Series
Mea
n =
Varia
nce
= Co
varia
nce
=
Mea
n =
Varia
nce
= Co
varia
nce
=
Mea
n =
Varia
nce
= Co
varia
nce
=
Mea
n =
Varia
nce
= Co
varia
nce
=
Mea
n =
Varia
nce
= Co
varia
nce
=
![Page 8: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/8.jpg)
Dr. M. Jahanur Rahman, R.U. 8
Ergodicity
a realization(a sample path)
Mea
n =
Varia
nce
= Co
varia
nce
=
Mea
n =
Varia
nce
= Co
varia
nce
=
Mea
n =
Varia
nce
= Co
varia
nce
=
Mea
n =
Varia
nce
= Co
varia
nce
=
Mea
n =
Varia
nce
= Co
varia
nce
=
![Page 9: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/9.jpg)
Dr. M. Jahanur Rahman, R.U.
Mea
n =
Varia
nce
= Co
varia
nce
=
Mea
n =
Varia
nce
= Co
varia
nce
=
Mea
n =
Varia
nce
= Co
varia
nce
=
Mea
n =
Varia
nce
= Co
varia
nce
=
Mea
n =
Varia
nce
= Co
varia
nce
=
Covariance = Variance =
Mean =
The covariance-stationary process is ergodic for the mean.
The covariance-stationary process is ergodic for the variance.
The covariance-stationary process is ergodic for the covariance.9
Ergodicity
![Page 10: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/10.jpg)
Dr. M. Jahanur Rahman, R.U. 10
Non-stationary time series Non-stationary time series
-3
-2
-1
0
1
2
3
10 20 30 40 50 60 70 80 90 00
Stationary time series
(Non)Stationay Time Series
![Page 11: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/11.jpg)
Dr. M. Jahanur Rahman, R.U. 11
Non-stationary time series contain two kinds of trends: Deterministic trend Stochastic (random) trend
Non-stationary Time Series
![Page 12: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/12.jpg)
Dr. M. Jahanur Rahman, R.U. 12
Trend Stationary: If y(t) contains a deterministic trend and
{y(t) β trend}
becomes stationary. Then y(t) is known as trend stationary.
Difference Stationary: If y(t) contains a stochastic trend and
{y(t) β y(t-1)}
becomes stationary. Then y(t) is known as difference stationary.
Non-stationary Time Series
![Page 13: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/13.jpg)
Dr. M. Jahanur Rahman, R.U. 13
Non-stationary Time Series
![Page 14: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/14.jpg)
Dr. M. Jahanur Rahman, R.U. 14
Any time series can contain some or all of the following
components:
1. Trend (T)
2. Cyclical (C)
3. Seasonal (S)
4. Irregular (I)
Trend component
The trend is the long term pattern of a time series. A trend can be
positive or negative depending on whether the time series exhibits an
increasing long term pattern or a decreasing long term pattern.
Components of a Time Series
![Page 15: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/15.jpg)
Dr. M. Jahanur Rahman, R.U. 15
Cyclical component
Any pattern showing an up and down movement around a given trend
is identified as a cyclical pattern. The duration of a cycle depends on
the type of business or industry being analyzed.
Any time series can contain some or all of the following
components:
1. Trend (T)
2. Cyclical (C)
3. Seasonal (S)
4. Irregular (I)
Components of a Time Series
![Page 16: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/16.jpg)
Seasonal component
Seasonality occurs when the time series exhibits regular fluctuations
during the same month (or months) every year, or during the same
quarter every year. For instance, retail sales peak during the month of
December.
Any time series can contain some or all of the following
components:
1. Trend (T)
2. Cyclical (C)
3. Seasonal (S)
4. Irregular (I)
Components of a Time Series
Dr. M. Jahanur Rahman, R.U. 16
![Page 17: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/17.jpg)
Dr. M. Jahanur Rahman, R.U. 17
Irregular component
This component is unpredictable. Every time series has some
unpredictable component that makes it a random variable. In prediction,
the objective is to βmodel" all the components to the point that the only
component that remains unexplained is the random component.
Any time series can contain some or all of the following
components:
1. Trend (T)
2. Cyclical (C)
3. Seasonal (S)
4. Irregular (I)
Components of a Time Series
![Page 18: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/18.jpg)
Dr. M. Jahanur Rahman, R.U. 18
Components of a Time Series
![Page 19: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/19.jpg)
Dr. M. Jahanur Rahman, R.U. 19
Two DomainsTraditionally, there are two ways to analyze time series data:
Time domain analysis
Time domain analysis examines how a time series process evolves
through time, with tools such as autocorrelation function.
Frequency domain analysis
Frequency domain analysis, also known as spectral analysis,
studies how periodic components at different frequencies
describe the evolution of a time series.
![Page 20: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/20.jpg)
Dr. M. Jahanur Rahman, R.U. 20
Time
Time domainFrequency
Frequency domain
Two Domains
![Page 21: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/21.jpg)
Dr. M. Jahanur Rahman, R.U. 21
The Autoregressive Integrated Moving Average (ARIMA)
models, or Box-Jenkins models, are a class of linear models
that is capable of representing stationary as well as non-
stationary time series.
Time Series Model
![Page 22: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/22.jpg)
Dr. M. Jahanur Rahman, R.U. 22
ARIMA(p,d,q) Model
βππ π‘=π+π1βππ π‘β 1+π2β
ππ π‘ β2+β―ππ βππ π‘βπ+ππ‘+π1ππ‘ β1+π2ππ‘β 2+β―+ππππ‘ βπ
= Response (dependent) variable at time
= Respose variable at time lags
= difference operator ()
= White noise error term
= Errors in previous time periods
= parameters
= Number of autoregressive terms
= Number of moving average terms.
![Page 23: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/23.jpg)
Dr. M. Jahanur Rahman, R.U. 23
ARIMA(p,0,q) = ARMA(p, q) Model
ARIMA(p,0,0) = AR(p) Model
ARIMA(0,0,q) = MA(q) Model
![Page 24: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/24.jpg)
Dr. M. Jahanur Rahman, R.U. 24
ARIMA models rely heavily on both autocorrelation (AC) and
partial autocorrelation (PAC) patterns in time series data.
![Page 25: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/25.jpg)
Dr. M. Jahanur Rahman, R.U. 25
ARIMA(2,0,0) = AR(2)
AC PA
C
1 2 3 4 5 6 7 β¦β¦.. Lags 1 2 3 4 5 6 7 β¦β¦.. Lags
ARIMA(0,0,2) = MA(2)1 2 3 4 5 6 7 β¦β¦.. Lags 1 2 3 4 5 6 7 β¦β¦.. Lags
AC
PAC
ARIMA(2,0,2) = ARMA(2, 2)
AC
PAC
1 2 3 4 5 6 7 β¦β¦.. Lags1 2 3 4 5 6 7 β¦β¦.. Lags
Theoretical pattern of ACF and PACFACF 0PACF = 0 for lag > 2
ACF = 0 for lag > 2; PACF 0
ACF PACF 0
![Page 26: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/26.jpg)
Dr. M. Jahanur Rahman, R.U. 26
Choosing an ARIMA model
ARIMA modeling requires a great deal of skill, which of course come from practice.
Model Pattern of ACF Pattern of PACFAR(p) Spikes decays exponentially. Significance spikes through lags p
MA(q) Significance spikes through lags q Spikes decays exponentially
ARMA(p,q) Spikes decays exponentially Spikes decays exponentially
Theoretical pattern of ACF and PACF
![Page 27: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/27.jpg)
Dr. M. Jahanur Rahman, R.U. 27
1. Identification of the model
(choosing tentative p, d, p for ARIMA)
2. Parameter estimation of the chosen model.
4. Forecasting
3. Diagnostic checking
(are the estimated residual white noise?)
yes
no
Box-Jenkins (BJ) Methodology
![Page 28: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/28.jpg)
Dr. M. Jahanur Rahman, R.U. 28
Steps of modeling a observed time series
with ARIMA
![Page 29: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/29.jpg)
Dr. M. Jahanur Rahman, R.U. 29
Original time series [
Log transformation
![Page 30: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/30.jpg)
Dr. M. Jahanur Rahman, R.U. 30Regular differencing
Seasonal differencing
![Page 31: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/31.jpg)
Dr. M. Jahanur Rahman, R.U. 31
Correlogram: Shows SAC & SPAC at different lags
![Page 32: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/32.jpg)
Dr. M. Jahanur Rahman, R.U. 32
ARIMA(2,0,0) = AR(2)
AC PA
C
1 2 3 4 5 6 7 β¦β¦.. Lags
1 2 3 4 5 6 7 β¦β¦..
ARIMA(0,0,2) = MA(2)1 2 3 4 5 6 7 β¦β¦.. Lags
1 2 3 4 5 6 7 β¦β¦..
AC
PAC
ARIMA(2,0,2) = ARMA(2, 2)
AC
PAC
1 2 3 4 5 6 7 β¦β¦1 2 3 4 5 6 7 β¦β¦.. Lags
Competitive Models:
(1) ARMA(2, 0) (2) ARMA(0, 1) (3) ARMA(2, 1)
Choosing an ARIMA model
![Page 33: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/33.jpg)
Dr. M. Jahanur Rahman, R.U. 33
Model MSE
ARMA(2, 0) 3521
ARMA(0, 1) 4019
ARMA(2, 1) 3998
Looking for the best Model
![Page 34: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/34.jpg)
Dr. M. Jahanur Rahman, R.U. 34
Finally Forecasting
![Page 35: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/35.jpg)
Frequency Domain Fourier Transformation Short-time Fourier Transformation Wavelet Transformation
![Page 36: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/36.jpg)
Dr. M. Jahanur Rahman, R.U. 36
Frequency Domain
To get further information from the time series that is not
readily available in the time domain.
Frequency Domain
Time Domain
![Page 37: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/37.jpg)
Dr. M. Jahanur Rahman, R.U. 37
Autocorrelation function or correlogram is used for analyzing time series in time domain.
rk
k
Correlogram
t
Yt
Periodic process with noise
Time Domain
Periodicities in data can be best determined by analyzing the time series in frequency domain.
Time series
![Page 38: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/38.jpg)
Dr. M. Jahanur Rahman, R.U. 38
Frequency Domain
Frequency Domain
Time Domain
Transformations:
Fourier Transformation (FT)
Short-time Fourier Transformation (SFT)
Wavelet Transformation (WT) β¦β¦β¦
![Page 39: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/39.jpg)
Dr. M. Jahanur Rahman, R.U. 39
Time series analysis in frequency domain is known as
frequency domain analysis or spectral analysis.
![Page 40: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/40.jpg)
Dr. M. Jahanur Rahman, R.U. 40
The Basics
= Amplitude (height of the function), R > 0
= Frequency (cycle/unit time)
= Phase (the starting point of the cosine function
which lies in )
(period).
π π‘=π πππ (2π ππ‘+π )
![Page 41: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/41.jpg)
Dr. M. Jahanur Rahman, R.U. 41
π π‘=π πππ (2π ππ‘+π )
t = running from 1 to 100. Red curve has , = 4/100 = 1/25 and Blue curve has , = 4/100 = 1/25 and Black curve has , = 4/100 = 1/25 and
The Basics
![Page 42: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/42.jpg)
Dr. M. Jahanur Rahman, R.U. 42
as
π π‘=π πππ (2π ππ‘+π )
The Basics
![Page 43: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/43.jpg)
Dr. M. Jahanur Rahman, R.U. 43
π¦ 2π‘=2 cos (2π 5 ππ‘ )+3sin (2π 5 ππ‘ )
π¦ 1π‘=2cos (2π 2 ππ‘ )+3sin (2π 2 ππ‘ )
π¦ π‘=π¦1 π‘+π¦2 π‘
Jean Baptiste Joseph Fourier(1768-1830)
The Basics
![Page 44: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/44.jpg)
Dr. M. Jahanur Rahman, R.U. 44
Mat
hem
atic
al C
once
pts
![Page 45: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/45.jpg)
Dr. M. Jahanur Rahman, R.U. 45
π (π₯ )= β¨ π (π₯) ,1 β©β¨1,1 β©
1+βπ=1
β β¨ π (π₯) ,cos (ππ₯) β©β¨cos (ππ₯ ) ,cos (ππ₯) β©
cos (ππ₯)+βπ=1
β β¨ π (π₯) , sin (ππ₯ )β©β¨sin (ππ₯ ), sin (ππ₯ )β©
sin (ππ₯ )
So π0=1
2π β«βπ
π
π (π₯ )ππ₯
ππ=1π β«
βπ
π
π (π₯ )cos (ππ₯ )ππ₯
ππ=1π β«
βπ
π
π (π₯ )π ππ (ππ₯ )ππ₯ π=1,2,3 ,β¦β¦
π (π₯ )=π0+βπ=1
β
ππcos (ππ₯ )+βπ=1
β
ππsin (ππ₯)
According to Fourier, any function in can be defined as
Fourier Series
![Page 46: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/46.jpg)
Dr. M. Jahanur Rahman, R.U. 46
or equivalently,
Where
for
Fourier Series
![Page 47: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/47.jpg)
Dr. M. Jahanur Rahman, R.U. 47
Time (t)
y(t)
Frequency (f)
Y(f)
Time domain to Frequency Domain
Time Domain Frequency Domain
![Page 48: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/48.jpg)
Dr. M. Jahanur Rahman, R.U. 48
π¦ 2π‘=2 cos (2π 5 ππ‘ )+3sin (2π 5 ππ‘ )
π¦ 1π‘=2cos (2π 2 ππ‘ )+3sin (2π 2 ππ‘ )
π¦ π‘=π¦1 π‘+π¦2 π‘
2f 5f
5f
2f
Time Domain Frequency Domain
Time domain to Frequency Domain
![Page 49: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/49.jpg)
Dr. M. Jahanur Rahman, R.U. 49
Discrete Fourier Transformation (DFT)Let a time sequence
and a frequency sequence
The DFT of is
Where
Note that and are of the same size.
The DFT of a time series is another time series.
![Page 50: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/50.jpg)
Dr. M. Jahanur Rahman, R.U. 50
The inverse DFT (IDET) of , is given by
Inverse Discrete Fourier Transformation (IDFT)
DFT
IDFT
3f 40f6f
Fourier Transformation
Inverse Fourier Transformation.
![Page 51: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/51.jpg)
Dr. M. Jahanur Rahman, R.U. 51
Spectral Density
Spectral density () is the amount of variance per interval of frequency: Angular frequency:
The plot of vs is called spectrum and as a whole called spectral density.
![Page 52: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/52.jpg)
Dr. M. Jahanur Rahman, R.U. 52
Spectral Density
πΌ (π )
ππ
Spectral density transforms information from time domain to frequency domain
Spectral density
![Page 53: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/53.jpg)
Dr. M. Jahanur Rahman, R.U. 53
Total area under the spectrum is equal to the variance of the process.
All information in frequency domain is extracted from the spectral density (or spectrum).
Spectral density () is the amount of variance per interval of frequency.A peak in the spectrum indicates an important contribution to
the variance at frequencies close the peak.Prominent spikes indicate periodicity. Several expressions for spectrum exist in literature.
Spectral Density
![Page 54: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/54.jpg)
Dr. M. Jahanur Rahman, R.U. 54
Spectral Density For a complete random series the spectral density function is
constant β termed as white noise.White noise indicates that no frequency interval contains
more variance than any other frequency intervals.
πΌ (π )
ππSpectral density of a White noise
![Page 55: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/55.jpg)
Dr. M. Jahanur Rahman, R.U. 55
3f 40f6f
DFT
Time Series Periodogram
IDFT
Filtered Series
Remove high
frequency component
ReconstructionDecomposition
Application of Fourier Transformation
![Page 56: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/56.jpg)
Dr. M. Jahanur Rahman, R.U. 56
0 0.2 0.4 0.6 0.8 1-3
-2
-1
0
1
2
3
0 5 10 15 20 250
100
200
300
400
500
600
All frequency components exist at all times
Stationary Time Series
Fourier Transformation
![Page 57: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/57.jpg)
Dr. M. Jahanur Rahman, R.U. 57
Frequency
Am
plitu
de
Fourier TransformationNon-stationary Series
![Page 58: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/58.jpg)
Dr. M. Jahanur Rahman, R.U. 58
0 0.5 1-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
0 5 10 15 20 250
50
100
150
Time Frequency0 0.5 1
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
0 5 10 15 20 250
50
100
150
Time Frequency
Different in Time Domain
Same in Frequency Domain
FT only gives frequency components exist in the seriesThe time and frequency Information can not be seen at the same
time which is required.
Fourier TransformationNon-stationary Series
![Page 59: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/59.jpg)
Dr. M. Jahanur Rahman, R.U. 59
Short-Time Fourier TransformationDennis Gabor (1946) In STFT, the signal is divided into small
enough segments, where these segments of the series can be assumed to be stationary.
For this purpose, a fixed window function "w" is chosen
![Page 60: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/60.jpg)
Dr. M. Jahanur Rahman, R.U. 60
Short-Time Fourier TransformationDennis Gabor (1946) In STFT, the signal is divided into small
enough segments, where these segments of the series can be assumed to be stationary.
For this purpose, a fixed window function "w" is chosen
![Page 61: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/61.jpg)
Dr. M. Jahanur Rahman, R.U. 61
FT X
Short-Time Fourier Transformation
![Page 62: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/62.jpg)
Dr. M. Jahanur Rahman, R.U. 62
FT X
Short-Time Fourier Transformation
![Page 63: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/63.jpg)
Dr. M. Jahanur Rahman, R.U. 63
FT X
Advantage & Disadvantage
Short-Time Fourier Transformation
![Page 64: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/64.jpg)
Dr. M. Jahanur Rahman, R.U. 64
FT X
Short-Time Fourier Transformation
![Page 65: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/65.jpg)
Dr. M. Jahanur Rahman, R.U. 65
FT X
Short-Time Fourier Transformation
![Page 66: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/66.jpg)
Dr. M. Jahanur Rahman, R.U. 66
Time stepFrequency
Am
plitu
deShort-Time Fourier Transformation
![Page 67: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/67.jpg)
Dr. M. Jahanur Rahman, R.U. 67
Time
Freq
uenc
y
Better time resolution; Poor frequency resolution
Better frequency resolution; Poor time resolution
Narrow windows give good time resolution, but poor frequency resolution.
Wide windows give good frequency resolution, but poor time resolution.
Advantage & Disadvantage
Short-Time Fourier Transformation
![Page 68: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/68.jpg)
Dr. M. Jahanur Rahman, R.U. 68
Wide windows do not provide good localization at high frequencies
Advantage & Disadvantage
Short-Time Fourier Transformation
![Page 69: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/69.jpg)
Dr. M. Jahanur Rahman, R.U. 69
Use narrower windows at high frequencies
Advantage & Disadvantage
Short-Time Fourier Transformation
![Page 70: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/70.jpg)
Dr. M. Jahanur Rahman, R.U. 70
Narrow windows do not provide good localization at low frequencies
Advantage & Disadvantage
Short-Time Fourier Transformation
![Page 71: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/71.jpg)
Dr. M. Jahanur Rahman, R.U. 71
Use wider windows at low frequencies
Advantage & Disadvantage
Short-Time Fourier Transformation
![Page 72: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/72.jpg)
Dr. M. Jahanur Rahman, R.U. 72
= Small value
(Wavelet coefficient)
.54321
TimeSc
ale
Scale = 1, time = 0
Changing Window at Multi-resolution Analysis
![Page 73: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/73.jpg)
Dr. M. Jahanur Rahman, R.U. 73
= Small value
(Wavelet coefficient)
.54321
TimeSc
ale
Scale = 1, time = 100
Changing Window at Multi-resolution Analysis
![Page 74: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/74.jpg)
Dr. M. Jahanur Rahman, R.U. 74
= Moderate value
(Wavelet coefficient)
.54321
TimeSc
ale
Scale = 1, time = 500
Changing Window at Multi-resolution Analysis
![Page 75: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/75.jpg)
Dr. M. Jahanur Rahman, R.U. 75
= Large value
(Wavelet coefficient)
.54321
TimeSc
ale
Scale = 1, time = 800
Changing Window at Multi-resolution Analysis
![Page 76: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/76.jpg)
Dr. M. Jahanur Rahman, R.U. 76
Mag
nitu
de
20 Hz 50 Hz 120 Hz
Changing Window at Multi-resolution Analysis
![Page 77: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/77.jpg)
Dr. M. Jahanur Rahman, R.U. 77
The Discrete Wavelet Transform (DWT)
π·ππ (π ,π )=βπ‘π (π‘ )Ξ¨ (π ,π )
= wavelet function = mother wavelet m = is location parameter n = scale parameter = time series
![Page 78: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/78.jpg)
Dr. M. Jahanur Rahman, R.U. 78
Wavelet families
![Page 79: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/79.jpg)
Dr. M. Jahanur Rahman, R.U. 79
Multi-level Decomposition of a Series with Wavelets
Time Series
cA3 cD3
cA2 cD2
cA0 cD0The decomposition tree can be schematically described as:
![Page 80: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/80.jpg)
Dr. M. Jahanur Rahman, R.U. 80
0 0 0 0 0 00 0 0 0 0 00 0 0 0 0 00 0 0 0 0 0 - 0 0 0 0 0 00 0 - 0 0 0 00 0 0 0 - 0 00 0 0 0 0 0 -
0 0 0 00 0 0 0
- - 0 0 0 00 0 0 - -
- - - -
Basis Constriction
![Page 81: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/81.jpg)
Dr. M. Jahanur Rahman, R.U. 81
Example
![Page 82: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/82.jpg)
Dr. M. Jahanur Rahman, R.U. 82
Multi-level Decomposition of a Series with Wavelets
Time Series1, 3, 5, 11, 12, 13, 0, 1
cA32.8, 11.3, 17.6, 0.7
cD3-1.4, -4.2, -0.7, -0.7
0 0 0 0 0 00 0 0 0 0 00 0 0 0 0 00 0 0 0 0 0 - 0 0 0 0 0 00 0 - 0 0 0 00 0 0 0 - 0 00 0 0 0 0 0 -
Level-1 orScale-1
![Page 83: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/83.jpg)
Dr. M. Jahanur Rahman, R.U. 83
Multi-level Decomposition of a Series with Wavelets
Time Series1, 3, 5, 11, 12, 13, 0, 1
cA32.8, 11.3, 17.6, 0.7
cD3-1.4, -4.2, -0.7, -0.7
Level-1 orScale-1
Wavelet coefficients = {2.8, 11.3, 17.6, 0.7, -1.4, -4.2, -0.7, -0.7}
![Page 84: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/84.jpg)
Dr. M. Jahanur Rahman, R.U. 84
Multi-level Decomposition of a Series with Wavelets
Time Series1, 3, 5, 11, 12, 13, 0, 1
cA32.8, 11.3, 17.6, 0.7
cD3-1.4, -4.2, -0.7, -0.7
cA210.0, 13.0
cD2-6.0, 12.0
Level-1 orScale-1
Level-2 orScale-2
0 0 0 00 0 0 0
- - 0 0 0 00 0 0 - -
Wavelet coefficients = {10.0, 13.0, -6.0, 12.0, -1.4, -4.2, -0.7, -0.7}
![Page 85: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/85.jpg)
Dr. M. Jahanur Rahman, R.U. 85
Multi-level Decomposition of a Series with Wavelets
Time Series1, 3, 5, 11, 12, 13, 0, 1
cA32.8, 11.3, 17.6, 0.7
cD3-1.4, -4.2, -0.7, -0.7
cA210.0, 13.0
cD2-6.0, 12.0
cA016.26
cD0-2.12
Level-1 orScale-1
Level-2 orScale-2
Level-3 orScale-3
- - - -
Wavelet coefficients = {16.2, -2.1, -6.0, 12.0, -1.4, -4.2, -0.7, -0.7}
![Page 86: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/86.jpg)
Dr. M. Jahanur Rahman, R.U. 86
Multi-level Decomposition of a Series with Wavelets
![Page 87: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/87.jpg)
Dr. M. Jahanur Rahman, R.U. 87
DWT Analysis
IDWT Reconstruction
Discrete Wavelet Transform (IDWT).
The input can be reconstructed using the Inverse
![Page 88: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/88.jpg)
Frequency Domain Application-1 Fourier Transformation
Spectral density
Determining periods
Removing Periodicities
Modeling and Forecasting
![Page 89: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/89.jpg)
Dr. M. Jahanur Rahman, R.U. 89
Autocorrelation function or correlogram is used for analyzing time series in time domain.
Time series rk
k
Correlogram
t
Yt
Periodic process with noise
Why Frequency Domain?
![Page 90: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/90.jpg)
Dr. M. Jahanur Rahman, R.U. 90
The steps of analyzing the data
Plot the time series
Plot the correlogram (to suspect the periodicity)
Plot the spectrum (to identify the periodicities)
Ik
wk
rk
k
Yk
t
![Page 91: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/91.jpg)
Dr. M. Jahanur Rahman, R.U. 91
The spectrum shows prominent spikes which represents the periodicities inherent in the data.
The period corresponding the any value of may be computed by
While the correlogram indicate the presence of periodicities in the data, the spectral analysis helps identify the significant periodicities themselves
Periodicities
![Page 92: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/92.jpg)
Dr. M. Jahanur Rahman, R.U. 92
Ik
wk
Ik
wk
Ik
wk
Ik
wk
Reconstruct new series by
removing the significance
periodicity one after another as
Removing Periodicities
![Page 93: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/93.jpg)
Dr. M. Jahanur Rahman, R.U. 93
The periodicities are tested for signoficance by defining the
following statistics (Kashyap and Rao 1976)
Where
Statistical significance of periodicities
![Page 94: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/94.jpg)
Dr. M. Jahanur Rahman, R.U. 94
Removing Periodicities
Significant periodicities are removed from the original time series to get a new series , where
The series containing the previous periodicities
where d is no. of periodicities removed (which are known to be significant)
The spectrum of new series is plotted and the spikes are observed until the series become random.
![Page 95: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/95.jpg)
Dr. M. Jahanur Rahman, R.U. 95
Example A monthly time series over the period 1987-2011 is used. Objective: Modeling and forecasting
Time Series ACF
Periodogram PACF
![Page 96: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/96.jpg)
Dr. M. Jahanur Rahman, R.U. 96
Spectral density shows tree significant periodic patterns with frequencies at
with corresponding periods 30 months 12 months 06 months respectively.
Correlogram & SpectrumCorrelogram indicates the presence
of periodicity.
ACF
Periodogram
![Page 97: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/97.jpg)
Dr. M. Jahanur Rahman, R.U. 97
One-period removed series
Two-period removed series
Three-period removed series
Correlogram & Spectrum
![Page 98: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/98.jpg)
Dr. M. Jahanur Rahman, R.U. 98
Model Selection
Seasonally differenced seriesARIMA(2,0,0) = AR(2)
Periodicities removed seriesARIMA(2,0,0) = AR(2)
PAC
PACAC
AC
![Page 99: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/99.jpg)
Dr. M. Jahanur Rahman, R.U. 99
Trail data: January, 1964 β December, 2007Test data: January, 2008 β December, 2008
Forecasting
![Page 100: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/100.jpg)
Dr. M. Jahanur Rahman, R.U. 100
Comparison of Forecasting PerformanceMean Absolute Error (MAE)Root Mean Squared Error (RMSE)Mean Absolute Percentage Error (MAPE)
Lower-value implies best forecast
![Page 101: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/101.jpg)
Dr. M. Jahanur Rahman, R.U. 101
seasonally differenced series Parameters are significant R-square is high DW-statistic is near 2
Periodicities removed series Parameters are significant R-square is moderate DW-statistic is about 2
EstimationAR(2) model
![Page 102: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/102.jpg)
Dr. M. Jahanur Rahman, R.U. 102
Forecasting
RMSE MAE PR series 1.173 4.739D series 0.852 3.914
![Page 103: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/103.jpg)
Dr. M. Jahanur Rahman, R.U. 103
Decomposition
Discrete Wavelet Transformation
Modeling & Forecasting
Humidity & Maximum Temperature of Rajshahi
Data obtained from BARC
Frequency Domain Application-2
![Page 104: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/104.jpg)
Dr. M. Jahanur Rahman, R.U. 104
Monthly Humidity of Rajshahi(1964-2008)
![Page 105: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/105.jpg)
Dr. M. Jahanur Rahman, R.U. 105
Humidity y(t)
A1
A2
A3
A4
D1
D2
D3
D4
Y = A1 + D1
Y = A2 + D2 +D1
Y = A3 + D3 + D2 +D1
Y = A4 + D4 + D3 + D2 +D1
Decom
positi
on by
wavele
t
Trans
form
ation
Monthly humidity of Rajshahi over 1964-2008
![Page 106: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/106.jpg)
Dr. M. Jahanur Rahman, R.U. 106
DecompositionLevel =3 Wavelet : Doubasis-5
Monthly humidity of Rajshahi over 1964-2008
![Page 107: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/107.jpg)
Dr. M. Jahanur Rahman, R.U. 107
Input series
Wavelet Transformatio
nForecasting
Inverse Wavelet
Transformation
Output series
ForecastingInput series
Wavelet Transformatio
n Processing
Inverse Wavelet
Transformation
Output series
Method-1
Method-2
Forecasting Framework
![Page 108: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/108.jpg)
Dr. M. Jahanur Rahman, R.U. 108
Forecasting Framework
Proposed by: Rocha, T. et.al. (2010)
![Page 109: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/109.jpg)
Dr. M. Jahanur Rahman, R.U. 109
Method-1Step-1: The wavelet transformation of type Daubechies-5 and decomposition level 3 is applied to the series ( result in 4 series denoted by and .
Step-2: Use a specific ARIMA model of each one of the constitutive series to forcast its future values, that is,
Step-3: The inverse wavelet transform is used to reconstruct the forecast series as
![Page 110: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/110.jpg)
Dr. M. Jahanur Rahman, R.U. 110
Method-2Step-1: The wavelet transformation of type Daubechies-5 and decomposition level 3 is applied to the series ( result in 4 series denoted by and .
Step-2: Reconstruct the series by removing the high frequency component
Step-3: Use the appropriate ARIMA model to the reconstructed series to forecast the future values;
![Page 111: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/111.jpg)
Dr. M. Jahanur Rahman, R.U. 111
Trail data: January, 1964 β December, 2007
Test data: January, 2008 β December, 2008
Forecasting of Humidity
![Page 112: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/112.jpg)
Dr. M. Jahanur Rahman, R.U. 112
Comparison of Forecasting PerformanceMean Absolute Error (MAE)Root Mean Squared Error (RMSE)Mean Absolute Percentage Error (MAPE)
Lower-value implies best forecast
![Page 113: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/113.jpg)
Dr. M. Jahanur Rahman, R.U. 113
Method MAE RMSE MAPE
Time domain Method 0.1633 0.2187 18.566
Frequency Domain Method-1
0.1336 0.1591 15.3654
Frequency Domain Method-2
0.0654 0.0799 9.1472
Forecasting performance over Jan-2008 to Dec-2008
Forecasting of HumidityModels
Time Domain: ARIMA(0,1,1)(0,1,1)
Frequency Domain-1: ARIMA(8,2,8)(0,0,1), ARIMA(0,1,2)(0,1,1),
ARIMA(0,0,4)(0,1,0), ARIMA(2,0,2)(0,0,0)
Frequency Domain-2: ARIMA(2,1,1)(0,1,1)
![Page 114: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/114.jpg)
Dr. M. Jahanur Rahman, R.U. 114
Frequency Domain Method-2 Time Domain Method
Method MAE RMSE MAPE
Time domain Method 0.1633 0.2187 18.566
Frequency Domain Method-1
0.16835 0.2296 19.001
Frequency Domain Method-2
0.0654 0.0799 9.1472
Forecasting of Humidity
![Page 115: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/115.jpg)
Dr. M. Jahanur Rahman, R.U. 115
Monthly Maximum Temperature of Rajshahi(1964-2009)
Trail data: Jan, 1964 β Dec, 2008
Test data: Jan, 2009 β Dec, 2009
![Page 116: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/116.jpg)
Dr. M. Jahanur Rahman, R.U. 116
Forecasting of Maximum Temperature
Monthly humidity of Rajshahi over 1964-2008
Method MAE RMSE MAPE
Time domain Method 0.7339 1.0507 2.2315
Frequency Domain Method-1
Frequency Domain Method-2
0.6995 0.8286 2.1402
Models
Time Domain: ARIMA(0,0,1)(0,1,1)
Frequency Domain-1: ARIMA(1,2,8)(0,0,1), ARIMA(4,1,4)(0,1,1),
ARIMA(0,0,6)(0,1,1), ARIMA(1,0,4)(0,1,1)
Frequency Domain-2: ARIMA(0,0,1)(0,1,1)
![Page 117: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/117.jpg)
Dr. M. Jahanur Rahman, R.U. 117
Forecasting of Maximum Temperature
Method MAE RMSE MAPE
Time domain Method 0.7339 1.0507 2.2315
Frequency Domain Method-1
Frequency Domain Method-2
0.6995 0.8286 2.1402
![Page 118: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/118.jpg)
Dr. M. Jahanur Rahman, R.U. 118
In modeling time series both time domain and
frequency domain analysis should be done
together.
For high frequency and noisy data, frequency
domain analysis performs better.
Recommendation
![Page 119: Frequency Domain Techniques i n Forecasting with ARIMA Model](https://reader036.vdocument.in/reader036/viewer/2022081418/5681625d550346895dd2b6a5/html5/thumbnails/119.jpg)
Dr. M. Jahanur Rahman, R.U. 119
Than
ks