Professional Documents
Culture Documents
1
Exponential Smoothing
Ft 1 Ft α( At Ft )
Ft 1 Ft αAt αFt
Ft 1 (1 α) Ft αAt
2
1-2. Short Questions and Multiple Choices
1. Given an actual demand of 60 for a period when forecast of 70
was anticipated, and an alpha of 0.3, what would the forecast for
the next period be using simple exponential smoothing?
F = (1-0.3)(70)+0.3(60) = 67
F(t+1) = Ft + (At-Ft)
65 66 +5
-5
65 = 66 + (-5) 5=1 = 0.2
5
6. Short Questions and Multiple Choices
6. The president of State University wants to forecast student
enrollments for the next academic year based on the following
historical data for the past 5 years. What is your forecast for next
year using exponential smoothing with α = 0.4?
α= 0.4
t At Ft |E| MAD
1 15000 15000 =B3 0 0
2 16000 15000 =(1-$C$1)*C3+$C$1*B3 1000 500
3 18000 15400 =(1-$C$1)*C4+$C$1*B4 2600 1200
4 20000 16440 =(1-$C$1)*C5+$C$1*B5 3560 1790
5 21000 17864 =(1-$C$1)*C6+$C$1*B6 3136 2059.2
19118.4 =(1-$C$1)*C7+$C$1*B7 2574
7
9-10. Short Questions and Multiple Choices
9. A forecast based on the previous forecast plus a percentage of
the forecast error is:
A) a naive forecast
B) a simple moving average forecast
F = (1-α)Ft+ α(At)
C) a centered moving average forecast t+1
D) an exponentially smoothed forecast Ft+1 = Ft+ α(At-Ft )
E) an associative forecast
8
Lecture of Problems 11a &11b.
https://www.youtube.com/watch?v=2OkoBbEm10I&t=9s
9
11a. Exponential Smoothing-Alpha using Data Table
α= 0.4 0 0.1
t At Ft E |E| MAD TS 1.25
1 19 19.00 0.00 0.00 0.00 #DIV/0! 0 1.28571
2 18 19.00 -1.00 1.00 0.50 -2.00 0.1 1.24839
3 21 18.60 2.40 2.40 1.13 1.24 0.2 1.22107
4 20 19.56 0.44 0.44 0.96 1.92 0.3 1.20159
5 18 19.74 -1.74 1.74 1.12 0.09 0.4 1.25134
6 22 19.04 2.96 2.96 1.42 2.15 0.5 1.30804
7 20 20.22 -0.22 0.22 1.25 2.27 0.6 1.36398
0.7 1.47358
0.8 1.59913
0.9 1.7278
1 1.85714
10
11b. Exponential Smoothing-Find Alpha Using Solver
α= 0.4
t At Ft E |E| MAD
1 19 19.00 0.00 0.00 0.00
2 18 19.00 -1.00 1.00 0.50
3 21 18.60 2.40 2.40 1.13
4 20 19.56 0.44 0.44 0.96
5 18 19.74 -1.74 1.74 1.12
6 22 19.04 2.96 2.96 1.42
7 20 20.22 -0.22 0.22 1.25
α= 0.31111
t At Ft E |E| MAD
1 19 19.00 0.00 0.00 0.00
2 18 19.00 -1.00 1.00 0.50
3 21 18.69 2.31 2.31 1.10
4 20 19.41 0.59 0.59 0.98
5 18 19.59 -1.59 1.59 1.10
6 22 19.10 2.90 2.90 1.40
7 20 20.00 0.00 0.00 1.20
11
12. Short Questions and Multiple Choices
11. It can be mathematically proved that the Age of data in
exponential smoothing is 1/α . The larger the α, the larger the
number of periods in the moving average. True or false? Why?
False
α = 0.5 age of data is 2 periods.
α = 0.2 age of data is 5 periods.
α = 0.1 age of data is 10 periods.
12
13-14. Short Questions and Multiple Choices
13. Given a forecast using a 6 period moving average. What is the
average age of data?
The last piece (newest piece) of data is only 1 period old.
The first piece (oldest piece) of data in a 6 period moving average
is 6 periods old.
The average age of data is (1+6)/2 = 3.5
13
15. Problem
Given the following demand data
Month Feb Mar Apr May Jun Jul Aug
Demand 19 18 15 20 18 22 20
At
25
15
2 18 19.0
3 15 18.8
4 20 18.0
5 18 18.4
6 22 18.3 18.0 4 4 3.65 3.65
7 20 19.1 18.6 1.4 2.7 0.92 2.29
19.3 19.0
15
15. Problem
d) Provide forecasts for September using Naïve method (forecast
of the next period is the same as the actual for this period).
Based on MAD, which method is better? ES or NM?
α= 0.2
t At Ft (ES) Ft(NM) |E|(NM) MAD(NM) |E|(ES) MAD(ES)
1 19 19.0
2 18 19.0 19 1 1 1.0 1.00
3 15 18.8 18 3 2 3.8 2.40
4 20 18.0 15 5 3 2.0 2.25
5 18 18.4 20 2 2.75 0.4 1.80
6 22 18.3 18 4 3 3.7 2.17
7 20 19.1 22 2 2.83 0.9 1.96
19.3 20
16
15. Problem
e) Provide forecasts and standard deviation of forecast for
September using linear Regression. Also Compute R-Squared
(Coefficient of determination) and Correlation Coefficient.
t At
1 19 17.4 Intercept 16.86 =INTERCEPT($B$3:$B$9,$A$3:$A$9)
2 18 17.9 Slope 0.5 =SLOPE($B$3:$B$9,$A$3:$A$9)
3 15 18.4 R-Squared 0.24 =RSQ($B$3:$B$9,$A$3:$A$9)
4 20 18.9 CC 0.49 =CORREL($B$3:$B$9,$A$3:$A$9)
5 18 19.4 StdDevOfForc 2.1 =STEYX($B$3:$B$9,$A$3:$A$9)
6 22 19.9 F8 20.86
7 20 20.4
8 20.9
9 21.4
10 21.9
11 22.4
17
(f) Which Technique ?
When comparing several methods, we need to use the same
time horizon for all methods. We need to have actual as well as
forecasts for all methods for all periods of MAD computations
Here we have Actual for periods 1 to 7; that is 7 periods.
Regression can provide us with forecast for periods 1 to ∞
Five period moving average can only provide forecast for
periods 6 and 7; that is 2 periods
Therefore, to compare all these methods, we can compute MAD
only over 2 periods. But two periods is not enough.
18