Forecasting Methods

Lecture 3
Time Series Models
Time Series Models
Naïve
Simple Moving Average
Weighted Moving Average
Single Exponential Smoothing
Adaptive Moving Average
Double Moving Average
Double Exponential Smoothing
Triple Exponential Smoothing
Time Series Models
Akua’s Bakery has been forecasting by ‘gut feel’. They
would like to use a formal forecasting technique. (i.e.
quantitative)
Exponential Smoothing(SES and AES)
SIMPLE EXPONENTIAL SMOOTHING
METHOD
This exponential smoothing method only requires that
you dig up two pieces of data to apply it
the most recent actual demand and
the most recent forecast.

An attractive feature of this method is that forecasts

made with this model will include a portion of every
piece of historical demand.
SIMPLE EXPONENTIAL SMOOTHING METHOD
This method is used when data pattern is
approximately horizontal (i.e., there is no neither
cyclic variation, nor seasonality nor pronounced trend
in the historical data
SIMPLE EXPONENTIAL SMOOTHING
METHOD
 The new forecast for next period (period t) will be
calculated as follows:
 F = F + α(A – F )
t t-1 t-1 t-1
Ft = αAt-1 + (1- α)Ft-1 (a bit more user friendly)

Where α is a smoothing coefficient whose value is between

0 and 1
SIMPLE EXPONENTIAL SMOOTHING
METHOD
EXPONENTIAL SMOOTHING
METHOD
Adaptive Exponential Smoothing
Adaptive exponential smoothing methods allow a
smoothing parameter to change over time, in order to
adapt to changes in the characteristics of the time
series.

The development of this method is an attempt to

overcome the fixed constant value of α by
incorporating the effect of the changing pattern of the
data series into the model
Adaptive Exp. Smoothing α=0.3, adaptive
= ±0.2
Month Actual Forecast Error α α+ α-
1 820
2 775
3 680
4 655
5 750
6 802
7 798
8 689
9 775
10 ???
Adaptive Exp. Smoothing α=0.3, adaptive
= ±0.2
Month Actual Forecast Error α α+ α-
1 820
2 775 820 -45
3 680 ?
4 655
5 750
6 802
7 798
8 689
9 775
10 ???
Adaptive Exp. Smoothing α=0.3, adaptive
= ±0.2
Month Actual Forecast Error α α+ α-
1 820
2 775 820 -45 0.3
3 680 ?
4 655
5 750
6 802
7 798
8 689
9 775
10 ???
Adaptive Exp. Smoothing α=0.3, adaptive
= ±0.2
Month Actual Forecast Error α α+ α-
1 820
2 775 820 -45 0.3
3 680 806.5 -126.5 ? 797.5 815.5
4 655
5 750
6 802
7 798
8 689
9 775
10 ???
Adaptive Exp. Smoothing α=0.3, adaptive
= ±0.2
Month Actual Forecast Error α α+ α-
1 820
2 775 820 -45 0.3
3 680 806.5 -126.5 0.5 797.5 815.5
4 655 ?
5 750
6 802
7 798
8 689
9 775
10 ???
Adaptive Exp. Smoothing α=0.3, adaptive
= ±0.2
Month Actual Forecast Error α α+ α-
1 820
2 775 820 -45 0.3
3 680 806.5 -126.5 0.5 797.5 815.5
4 655 743.25 -88.25 ? ?
5 750
6 802
7 798
8 689
9 775
10 ???
Adaptive Exp. Smoothing α=0.3, adaptive
= ±0.2
Month Actual Forecast Error α α+ α-
1 820
2 775 820 -45 0.3
3 680 806.5 -126.5 0.5 797.5 815.5
4 655 743.25 -88.25 ? 717.95 768.55
5 750
6 802
7 798
8 689
9 775
10 ???
Adaptive Exp. Smoothing α=0.3, adaptive
= ±0.2
Month Actual Forecast Error α α+ α-
1 820
2 775 820 -45 0.3
3 680 806.5 -126.5 0.5 797.5 815.5
4 655 743.25 -88.25 0.7 717.95 768.55
5 750
6 802
7 798
8 689
9 775
10 ???
AES with α = 0.3 and adaptive α = +- 0.2
What if there is trend in the data?
What is Double Moving Average?
Applies the moving average technique twice, once to
the original data and then to the resulting single
moving average data.

This method then uses both sets of smoothed data to

project forward.

This method is best for historical data with a trend but

no seasonality.
 Double Moving Average
 SMA of n periods

 DMA of n periods

 Level at time t

 Slope at time t

 Forecast at time t+m

Double Moving Average
Double Exponential Smoothing
There are three types of exponential smoothing
methods used in Holt-Winters:
• Single Exponential Smoothing – suitable for
forecasting data with no trend or seasonal pattern,
where the level of the data may change over time.
• Double Exponential Smoothing – for forecasting data
where trends exist.
• Triple Exponential Smoothing – used for forecasting
data with trend and/or seasonality.
Double Exponential Smoothing
Also known as Holt linear trend method. Holt
(1957) extended simple exponential smoothing
to allow the forecasting of data with a trend.

St = αDt + (1 – α)(St-1 + Gt-1)

Gt = β(St – St-1) + (1 – β)Gt – 1
Ft + 1 = St + rGt
Double Exponential Smoothing
Triple Exponential Smoothing
 Triple Exponential Smoothing
Holt (1957) and Winters (1960) extended Holt’s method to
capture seasonality.
The Holt-Winters seasonal method comprises the forecast
equation and three smoothing equations:
 one for the level ℓt, one for the trend bt, and one for the seasonal
component st, with corresponding smoothing parameters α, β∗
and γ.
We use m to denote the frequency of the seasonality, i.e., the
number of seasons in a year. For example, for quarterly data
m=4, and for monthly data m=12.
The two main variations of Holt's Exponential Smoothing with
seasonal components are the multiplicative method and the
additive method.
 Triple Exponential Smoothing
 The additive method is preferred when the seasonal variations
are roughly constant through the series, while the
multiplicative method is preferred when the seasonal
variations are changing proportional to the level of the series.

 With the additive method, the seasonal component is

expressed in absolute terms in the scale of the observed
series, and in the level equation the series is seasonally
adjusted by subtracting the seasonal component. With the
multiplicative method, the seasonal component is expressed
in relative terms (percentages), and the series is seasonally
adjusted by dividing through by the seasonal component.
 Triple Exponential Method
 Holt's Multiplicative Exponential Smoothing:
In the multiplicative method, the seasonal component is multiplied
by the level and trend components. This means that the seasonal
effect is considered as a relative value or a percentage change
relative to the baseline.
Forecast (Ft) = (Level (lt) + Trend (bt)) * Seasonal (st)
Use the multiplicative method when the seasonal variation is
proportional to the level of the time series. For instance, if you are
forecasting daily website traffic, and the increase or decrease in
traffic during a certain season is proportional to the overall level of
traffic, then the multiplicative method might be more appropriate.
Within each year, the seasonal component will sum up to
approximately m.
Triple Exponential Method
Triple Exponential Method
 Triple Exponential Method
 Holt's Additive Exponential Smoothing:
In the additive method, the seasonal component is added to the level
and trend components. This means that the seasonal effect is
considered as an absolute value or an increment to the baseline.
Forecast (Ft) = Level (lt) + Trend (bt) + Seasonal (st)
Use the additive method when the seasonal variation remains roughly
constant in magnitude regardless of the level of the time series. For
example, if you are forecasting monthly sales, and the increase or
decrease in sales during a certain season is relatively constant in
absolute terms, then the additive method might be appropriate.
Within each year, the seasonal component will add up to
approximately zero.
Triple Exponential Method
 Take Note:
 Each smoothed value is the weighted average of the previous observations, where the
weights decrease exponentially depending on the value of parameter (α)

 If α is equal to 1 (one) then the previous observations are ignored entirely

 if α is equal to 0 (zero), then the current observation is ignored entirely, and the
smoothed value consists entirely of the previous smoothed value

 Values in between produce intermediate results

 Large values of α make a forecast more responsive to more recent levels

 Small values of α have a damping effect

 Beta constants have similar effects