Section 1 Forecasting

Forecasting -telling in advance a possible event that may take place

Methods
1. Extrapolation – assumes that historical data contains a stable pattern that will continue in
the next periods
2. Causal – attempts to find a relationship between one or more other variables
3. Judgmental – human factor is the only realistic method.
Time Series Patterns
Time-series – a series of data collected at regular intervals of time such as every week,
every month
1. Constant trend - no trend.
2. Linear Trend - straight line trend for any period in the future.
3. Exponential trend - amount of growth increases rapidly; unrealistic at long horizons
4. Damped trend – trend declines each period until it dies out;for long-range planning

1. 2. 3. 4.

Section 2 Forecasting Models

The most common forecasting models used in business are the Naïve, Moving Average,
Exponential Smoothing and Time-Series Regression.

1.Naive- the present data is used as a forecast for the next period.
Mathematical Model : Ft+1= Xt

Period(t Data(Xt Forecast(F

) ) t)
1 28
2 27 28
3 33 27
4 25 33
5 34 25
6 33 34
7 35 33
8 30 35
9 33 30
10 35 33
11 27 35
12 29 27
13 29
2. Moving Average- average of the most recent data
a. Unweighted MA b.) Weighted MA

Mathematical Model :
UMA: Ft+1=(Xt+Xt-1+Xt-2)/3 3-period unweighted moving
average
Ft+1=(Xt+Xt-1+Xt-2+Xt-3)/4 4-period unweighted
WMA: Ft+1=(3Xt+2Xt-1+1Xt-2)/6 3-period weighted moving average
Ft+1=(4Xt+3Xt-1+2Xt-2+1Xt-3)/10 4-period weighted

Unweighted MA Weighted MA
Period(t) Data(Xt) Forecast

1 28 F
2 27 4 F
3 33
5
4 25 F4=(33+27+28)/3= 29.33 F4 =3(33)+2(27)+1(28)/6= 30.17
5 34 F5=(25+33+27)/3= 28.33 F5 =3(25)+2(33)+1(27)/6= 28.00
6 33 30.67 30.83
7 35 30.67 32.00
8 30 34.00 34.17
9 33 32.67 32.17
10 35 32.67 32.33
11 27 32.67 33.50
12 29 31.67 30.67
13 30.33 29.33

3.Simple Exponential Smoothing

-uses a single weighting factor
- the new forecast is equal to the old forecast plus a fraction of the error.
Mathematical Model: Ft+1=Ft + et where: Ft = Forecast for t (present period)
et = error in t (et=Xt-Ft); forecast error
= fraction of the error; between 0 and 1
Simple Exponential Smoothing with F1=30 and  =0.10
Period(t) Data(Xt) Forecast(Ft) Error (et) Ft+1=Ft + et
1 28 30 28-30=-2 F2=30+0.10(-2)= 29.8
2 27 29.8 27-29.8=-2.8 F3=29.8+0.10(-2.8)= 29.52
3 33 29.52 33-29.52=3.48 F4=29.52+0.10(3.48)= 29.868
4 25 29.868 29.3812
5 34 29.3812 29.84308
6 33 29.84308 30.158772
7 35 30.158772 30.6428948
8 30 30.6428948 30.57860532
 9 33 30.5786053 30.82074479
10 35 30.8207448 31.23867031
11 27 31.2386703 30.81480328
12 29 30.8148033 30.63332295
13 30.63332295

4.Trend Projection(Time-Series Regression)

-fits a trend line to a set of past observations projecting the line into the future periods
Mathematical Model : Ft = a + bt where : a = intercept
( ∑ tX-n t ' X' )
b= a = X' – bt' b = slope
( ∑ t 2 -nt’ 2 )
X= dependent variable(usually sales)
t = time (period)
∑t
t' = = mean of values of t
n
∑X
X' = = mean of values of X
n
Time-Series Regression(Trend Projection)
Period(t) Data(X) tX t2
(Jan)1 28 28 1
(Feb)2 27 54 4
(Mar)3 33 99 9
(Apr)4 25 100 16
(May)5 34 170 25
(Jun)6 33 198 36
(Jul)7 35 245 49
(Aug)8 30 240 64
(Sep)9 33 297 81
(Oct)10 35 350 100
(Nov)11 27 297 121
(Dec)12 29 348 144
n=12 ∑t=78 ∑X=369 ∑tX=2426 ∑t2=650
t'=
78
=¿
12 6.5
X'=
369
=¿
12 30.75
[ 2426 -12(6.5)(30.75)]
b=
[ 650-12(6.5)2 ]
=0.192
a= 30.75-0.192(6.5) =29.5
Ft= 29.5+0.192 t  working equation(model)
When t =1 , F1 =29.5+0.192(1) = 29.69
t =2 , F2 =29.5+0.192(2) = 29.88
 t =3 , F3 =29.5+0.192(3) = 30.08
:
:
t =12 , F12 =29.5+0.192(12) = 31.80
What is the forecast for June next year?
Since June next year is the 18th period, then F18 = 29.5+0.192(18)= 32.96