Scribd
Upload
104 views

Exponential Smoothing

Exponential smoothing is a simple forecasting technique that places declining weights on past observations over time. It involves fitting a trend line to data points where the forecast is the current forecast plus a fraction (w) of the current forecast error. The method places geometrically declining weights on past values of the data series, with a large w value placing more weight on recent observations. The Holt-Winters method extends this to include trend and seasonal components.

Uploaded by

ratan203
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
104 views

Exponential Smoothing

Exponential smoothing is a simple forecasting technique that places declining weights on past observations over time. It involves fitting a trend line to data points where the forecast is the current forecast plus a fraction (w) of the current forecast error. The method places geometrically declining weights on past values of the data series, with a large w value placing more weight on recent observations. The Holt-Winters method extends this to include trend and seasonal components.

Uploaded by

ratan203
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
You are on page 1/ 5

Exponential Smoothing

All statistical fitting methods involve some type of


smoothing
 e.g., in regression, a trend line is fitter to the data points
Consider the following forecasting scheme
 ft = wyt-1 + (1-w)ft-1
 or, if you update by one period
 ft+1 = wyt + (1-w)ft

Another way to write the process is


 ft+1 = ft + w(yt – ft)
 The forecast is the current forecast plus a fraction of the
current forecast error.

Exponential Smoothing
Weighting Past Values

Exponential smoothing places declining weights on


past values of a series. Consider
ft+1 = wyt + (1-w)ft
= wyt + (1-w)[wyt-1 + (1-w)ft-1 ]
= w[yt + (1-w)yt-1] + (1-w)2(ft-1)
= w[yt + (1-w)yt-1] + (1-w)2[wyt-2 + (1-w)ft-2 ]
= w[yt + (1-w)yt-1+(1-w)2yt-2 ] + (1-w)3ft-3

f t 1  w (1  w)i yt i
i 0
Exponential smoothing places geometrically declining weights
on past valued of {yt}

Exponential Smoothing
Properties of the weights

Note that w must be


between 0 and 1 Period Value of i (1 - w)^i w*(1-w)^i
A large value of w places a t 0 1 0.8
high weight on the current t-1 1 0.2 0.16
realization of the series t-2 2 0.04 0.032
The sum of the weights on t-3 3 0.008 0.0064
past values of {yt} is unity. t-4 4 0.0016 0.00128
 The spreadsheet shows a t-5 5 0.00032 0.000256
value of w = 0.8
t-6 6 6.4E-05 5.12E-05
SUM = 0.999987

Exponential Smoothing
Holt-Winters

Other smoothing methods involve smoothing a trend lines fit to a


data series
ft+1 = wyt + (1-w)(ft+Tt)
Tt+1 = b(ft+1 – ft) +(1-b)Tt
The one-step ahead forecast is
Ht+1 = Ft+1 + Tt+1
The m=step ahead forecast is
 Ht+m = Ft+1 + mTt+1
Winter’s method includes seasonals
In forecast x you select
 alpha for the level
 beta for the seasonal
 gamma for the trend

Exponential Smoothing
Hints

The moving average and exponential smoothing methods will


work well with series that do not have a trend or structural
breaks.
 Exponential smoothing will be preferable if you want to heavily
weight near-term events.
Use the Holt-Winters Method for data that has trend and
seasonals.
 Select values alpha, beta and gamma near 1 if you want forecasts
to change strongly with new information.
 Forecast-x will also select the smoothing coefficients for you.
The Mean Square Forecast Error (or RMSE) is often used a a
criterion as to the best smoothing constant.
The adaptive-response model is limited; it will not be discussed
in class

Exponential Smoothing

You might also like