08-10-2020

Important points to remember:

• Simple Moving Average
• Weighted moving average
Tutorial No.: 1 • Exponential Smoothing
Operations and Planning Control

Simple moving Simple moving

average average
Useful if we can assume that market demands will stay fairly
steady over time

 Formula: Moving average = ∑demand in previous n periods

N
*Where: n – is the number of periods in the moving
average

10/8/2020 3

Simple moving
average Weighted moving
Example: compute a three-period moving average forecast given the following
demand for cars for the last five periods.
average
peirod Supply
1 70 Uses an average of a specified number of the most recent observations,
2 80 with each observation receiving a different emphasis (weight) when
3 65
there is a trend or pattern
4 90
5 85
 Formula: weighted moving average = ∑[(weight for period n) (demand in period n)]
• Solution: the forecast for period 6 should be:
∑Weights
Moving average forecast = 65 + 90 + 85 = 80 cars
3
If the actual demand in period 6 turns out to be 95, the moving average forecast for the
period 7 would be:

Moving average forecast = 90 + 85 + 95 = 90 cars

3
10/8/2020 5 10/8/2020 6

1
 08-10-2020

Weighted moving
average Weighted moving
average
• Example: compute a three- period weighted moving average forecast
given the following demand for cars the last five periods; with an
assigned weight of 1,2,3.

period demand
1 70
2 80
3 65
4 90
5 85

10/8/2020 8

Weighted moving Exponential

average smoothing
Solution: the forecast for period 6 would be: Used to forecast sales when there is no pattern in the demand for goods or
services
Fw= 65(1)+90(2)+85(3) = 65+180+245 = 490/6 = 81.67 or 82 cars
Formula: Ft = Ft-1 + ∝ [ at-1 – Ft-1]
(1+2+3) 6
*Where: Ft = new forecast or forecast for period
If the actual demand in period 6 turns out to be 95, the weighted ft-1 = previous forecast
moving average forecast for period 7 would be: ∝ = Smoothing constant

Wma = 90(1)+85(2)+95(3)/(3+2+1)
At-1 = actual demand or sales for period t-1
= (90+170+85)/6

= 545/6 = 90.83 or 91cars

10/8/2020 9 10/8/2020 10

Exponential Errors
smoothing
• Example 1: A car dealer predicted a January
demand for 550 Honda v-tech cars. Actual January Sum of absolute
demand was 680 Honda v-tech cars and Deviation (SAD)
• ∝ = 0.10. Forecast the demand for January, using Exp. ∝ = 0.10. Forecast the
the exponential smoothing model. demand and calculate the error
SAD = ∑ |Et |
period demand
Solution:
Ft = ft-1 + ∝ [ at-1 – ft-1]
1 70
= 550 + 0.10 [680-550]
2 80
= 550 + 0.10 [130] | Et |= | actual – forecast |
3 65
= 550 + 13
Ft = 563
10/8/2020 11

2
 08-10-2020

period demand forecast error Absolute MEAN ABSOLUTE ERROR

error
1 70 70 0 0
2 80 70 10 10
3 65 71 -6 6
∑ |Actual − Forecast|
MAD =
Ft = ft-1 + ∝ [ at-1 – ft-1]
= 70 + 0.10 [0] SAD = ∑ |Et |
period demand forecast error Absolute
Ft = 70 error
1 70 70 0 0
2 80 70 10 10
Ft = ft-1 + ∝ [ at-1 – ft-1] SAD=(0+10+6) 3 65 71 -6 6

= 70 + 0.10 [10] 16 MAD (α =0.1)

Ft = 71 =16/3 = 5.33

MAPE(Mean Absolute Percent Error)

MEAN SQUARE ERROR

∑ 100|(Actual − forecast ) |
MAPE =
Actual

period demand forecast error Absolute Absolute

error error /Dt
1 70 70 0 0 0
period demand forecast error Absolute (error)2
2 80 70 10 10 .125
error
3 65 71 -6 6 0.92 (0+100+36)
1 70 70 0 0 0
MSE=
=45.33
∑ 100|0 + .125 + 0.92 | 2 80 70 10 10 100
= 7.23
3 3 65 71 -6 6 36

Q1. The monthly demand for motorcycles for last 14 months is given below. (1) Forecast for March
Month Demand Month Demand
(1000 units) (1000 units)

January 1.3 August 2.0

Moving average =∑demand in previous n periods
February 1.1 September 2.9
N
March 0.9 October 3.2
*Where: n is the number of periods in the moving
April 1.6 November 3.7
average
May 1.5 December 2.4
June 2.1 January 1.6
July 2.4 Februrary 1.0

(1) Develop a forecast for March using a) 3 Month moving average = (2.4+1.6+1)/3=1.67
i) A 3-month and 4-months moving average
ii) A 3- month weighted moving average with weights of 0.1, 0.3, 0.6 for the earliest to latest
period. b) 4 Month moving average = (3.7+2.4+1.6+1)/4=2.175
iii) Develop an exponentially smoothed forecast with  = 0.1 and 0.3

(2) Determine forecast errors, SAD, MAD, MAPE and MSE for the forecasting models used in (i)
and comment on the results

3
 08-10-2020

Month Demand Forecast Error (Et) Forecast

(1000 units) (α=0.1) (actual demand- error
b) Weighted moving average with k=3 and weights=0.1,0.3,0.6 forecast) (0.3)
January 1.3 1.3 0
Feb 1.1
weighted moving average = ∑[(weight for period n) (demand in period n)] March 0.9
∑Weights April 1.6

Fmarch= [ (2.4×0.1)+(1.6×0.3)+(1×0.6)] = 1.32

(0.1+0.3+0.6) Ft = Ft-1 + α (Dt-1 – Ft-1 )

Error = actual- forecast

Ffeb = Fjan + 0.1 (Djan – Fjan ) 1.1-1.3 = -0.2
1.3+ 0.1 (1.3-1.3) = 1.3

Month Demand Forecast Error (Et)

(1000 units) (α=0.1) (actual-
month Demand Forecast Error (Et) forecast)
(1000 units) (α=0.1) (actual- January 1.3 1.3 0
forecast)
Feb 1.1 1.3 -0.2
January 1.3 1.3 0
March 0.9 1.28 -0.38
Feb 1.1 1.3 -0.2
April 1.6
March 0.9
April 1.6

Ft = Ft-1 + α (Dt-1 – Ft-1 ) Error = actual- forecast

Ft = Ft-1 + α (Dt-1 – Ft-1 ) 1.6 -1.242 = 0.358
Fapril = Fmarch + 0.1 (Dmarch – Fmarch )
Fmarch = Ffeb + 0.1 (Dfeb – Ffeb ) Error = actual- forecast
0.9 -1.29 = -0.38 = 1.28+ 0.1 (-0.38) = 1.242
= 1.3+ 0.1 (1.1-1.3) = 1.28

Month Demand (1000 Forecast Error (Et) Month Demand Forecast Error (Et) Forecast
Units) (α=0.1)
(1000 units) (α=0.1) (actual- (0.3)
January 1.3 1.3 0 forecast)
February 1.1 1.3 -0.2 January 1.3 1.3 0 1.3
March 0.9 1.28 -0.38 Feb 1.1 1.3 -0.2
April 1.6 1.242 0.358 March 0.9 1.24 -0.34
May 1.5 1.28 0.22
April 1.6
June 2.1 1.30 0.8
July 2.4 1.38 1.02
Ffeb = Fjan + 0.3 (Djan – Fjan ) Fapril = Fmarch + 0.3 (Dmarch – Fmarch )
August 2.0 1.48 0.52
September 2.9 1.53 1.37 1.3+ 0.3 (1.3-1.3) = 1.3 = 1.24+ 0.3 (-0.34) = 1.138
October 3.2 1.66 1.54
November 3.7 1.81 1.89
December 2.4 1.99 0.41
January 1.6 2.03 -0.43 Fmarch = Ffeb + 0.3 (Dfeb – Ffeb )
February 1.0 1.98 -0.98
= 1.3+ 0.3 (1.1-1.3) = 1.24

4
 08-10-2020

Month Demand (1000 Forecast Error (Et)

Units) (α=0.3) (actual-forecast)
January 1.3 1.3 0 (c) α=0.1 , 0.3
February 1.1 1.3 -0.2
March 0.9 1.24 -0.34
April 1.6 1.14 0.46
Ft = Ft-1 + α (Dt-1 – Ft-1 )
May 1.5 1.28 0.22
June 2.1 1.35 0.75
July 2.4 1.57 0.83
August 2.0 1.81 0.18 Forecast for March (α =0.1) =1.98+0.1(1-1.98)
September 2.9 1.87 1.03 =1.88
October 3.2 2.18 1.02
November 3.7 2.48 1.22 Forecast for March (α =0.3) =2.37+0.3(1-2.37)
December
January
2.4
1.6
2.84
2.70
-0.44
-1.1
=1.96
February 1.0 2.37 -1.37

Month Demand (1000 Forecast Error (Et) Forecast Error (Et)

Units) (α=0.1) (α=0.3) (actual-
forecast) MAD(Mean Absolute Deviation) =
Sum ofJanuary
absolute1.3 1.3 0 1.3 0 Here n=14
Deviation
February 1.1 1.3 -0.2 1.3 -0.2
March 0.9 1.28 -0.38 1.24 -0.34
April
SAD = ∑ |Et |
1.6 1.242 0.358 1.14 0.46
MAD = ∑ |Actual − Forecast|
May 1.5 1.28 0.22 1.28 0.22
June 2.1 1.30 0.8 1.35 0.75
July 2.4 1.38 1.02 1.57 0.83
MAD (α =0.1) = 10.118/14 = 0.722
| Et August
|= | actual 2.0
– forecast | 1.48 0.52 1.81 0.18
September 2.9 1.53 1.37 1.87 1.03
October 3.2 1.66 1.54 2.18 1.02 MAD (α =0.3) = 9.16/14 = 0.634
November 3.7 1.81 1.89 2.48 1.22
SAD (α =0.1)2.4=10.118
December 1.99 0.41 2.84 -0.44
January 1.6 2.03 -0.43 2.70 -1.1
SAD (α =0.3)1.0= 9.16
February 1.98 -0.98 2.37 -1.37

Month Deman Forecas Error (Et) |Et/Dt| Forecast (α=0.3) Error (Et)
MAPE(Mean Absolute Percent Error) d (1000 t (α=0.1) (actual-
Units) (α=0.1) forecast)
Jan 1.3 1.3 0 0 1.3 ∑ 100| (Actual − forecast ) |
0 Actual
MAPE =
Feb 1.1 1.3 -0.2 0.18 1.3 -0.2
∑ 100|(Actual − forecast ) | March 0.9 1.28 -0.38 0.42 1.24 -0.34
MAPE =
Actual
April 1.6 1.242 0.358 0.22 1.14 0.46
May 1.5 1.28 0.22 0.15 1.28 0.22
June 2.1 1.30 0.8 0.38 MAPE
1.35 (α=0.1)= 4.91/14×100=35.07
0.75
July 2.4 1.38 1.02 0.42 1.57 0.83
Aug 2.0 1.48 0.52 0.26 1.81 0.18
Sep 2.9 1.53 1.37 0.47 1.87 1.03
October 3.2 1.66 1.54 0.48 2.18 1.02
Nov 3.7 1.81 1.89 0.51 2.48 1.22
Dec 2.4 1.99 0.41 0.17 2.84 -0.44
January 1.6 2.03 -0.43 0.27 2.70 -1.1
February 1.0 1.98 -0.98 0.98 2.37 -1.37

5
 08-10-2020

Month Demand Forecast Error Et2 Forecast Error (Et) Et2

Month Deman Foreca Error |Et/Dt| Forecast Error (Et) |Et/Dt| (1000 (α=0.1) (Et) (α=0.3) (actual-
d st (Et) (α=0.3) (actual-
(α=0.1) (α=0.3)
(α=0.1) (α=0.3) Units) forecast)
(1000 (α=0.1 forecast)
Units) ) January 1.3 1.3 0 0 1.3 0 0
Jan 1.3 1.3 0 0 1.3 0 0 Februar 1.1 1.3 -0.2 0.04 1.3 -0.2 0.04
y
Feb 1.1 1.3 -0.2 0.18 1.3 -0.2 0.18
March 0.9 1.28 -0.38 0.14 1.24 -0.34 0.11
March
∑
0.9 (Actual
1.28
100|1.242
-0.38− forecast
0.42 )
|
1.24 -0.34 0.37
April 1.6 1.242 0.358 0.13 1.14 0.46 0.21
April
MAPE = 1.6 Actual 0.22
0.358 1.14 0.46 0.29
May 1.5 1.28 0.22 0.15 1.28 0.22 0.15
May 1.5 1.28 0.22 0.05 1.28 0.22 0.05
June 2.1 1.30 0.8 0.38 1.35 0.75 0.36
June 2.1 1.30 0.8 0.64 1.35 0.75 0.56
July 2.4 1.38 1.02 0.42 1.57 0.83 0.34 July 2.4 1.38 1.02 1.04 1.57 0.83 0.69
Aug 2.0 1.48 0.52 0.26 1.81 0.18 0.09 August 2.0 1.48 0.52 0.27 1.81 0.18 0.03
=MAPE (α=0.3)= 5.02/14×100=35.55 Sep 2.9 1.53 1.37 1.87 1.87 1.03 1.06
Sep 2.9 1.53 1.37 0.47 1.87 1.03 0.35
October 3.2 1.66 1.54 0.48 2.18 1.02 0.32 October 3.2 1.66 1.54 2.37 2.18 1.02 1.04
Nov 3.7 1.81 1.89 0.51 2.48 1.22 0.33 Nov 3.7 1.81 1.89 3.57 2.48 1.22 1.45
Dec 2.4 1.99 0.41 0.17 2.84 -0.44 0.18 Dec 2.4 1.99 0.41 0.17 2.84 -0.44 0.19
January 1.6 2.03 -0.43 0.27 2.70 -1.1 0.69 January 1.6 2.03 -0.43 0.18 2.70 -1.1 1.21
Februar 1.0 1.98 -0.98 0.98 2.37 -1.37 1.37 Feb 1.0 1.98 -0.98 0.96 2.37 -1.37 1.87
y

Q2. The manager of an ice cream parlor needs an accurate forecast of

demand for ice cream. For this he collects demand data for last 10
weeks as given below
MSE(Mean Sum Of Square Error) = Week Liters Sold
1 110
2 95
3 135
4 160
5 97
6 105
7 120
8 175
9 140
10 121
MSE(α=0.1)= 11.43/14=0.816
A) Develop exponentially smoothed forecast with  = 0.2 and use it for
forecasting demand for 11th period.
MSE (α=0.3)= 8.55/14=0.16 B)Compute MAD, MAPE and MSE for the forecasting models
developed in (a)

Forecast for 11 week

Week Dt( Liter Ft (α=0.2) Et=(Dt-Ft) Et/Dt Et2 F11 = F10 + α (D10 – F10 )
sold) Forecast

1 110 125.8 -15.8 0.14 249.64

F11 = 132.17 + 0.2 (-11.17)=129.94
2 95 122.64 -27.64 0.29 763.97
3 135 117.11 17.88 0.13 319.69
Here n = 10
4 160 120.68 39.32 0.24 1546.06
5 97 128.54 -31.54 0.32 994.77
MAD=∑Et /n =213.62/10=21.36
6 105 122.23 -17.23 0.16 296.87
7 120 118.78 1.22 0.01 1.48
MAPE = ∑(Et /Dt )/Dt)*100=1.77/10*100=17.7
8 175 119.02 55.98 0.32 3133.76
9 140 130.21 9.97 0.07 95.84
MSE = ∑Et2 / n=7526.84/10=752.684
10 121 132.17 -11.17 0.09 124.76

6
 08-10-2020

EXPONENTIAL SMOOTHING WITH

EXPONENTIAL SMOOTHING WITH TREND ADJUSTMENT
TREND ADJUSTMENT
Ft = exponentially smoothed forecast of the data series in
period t

EXPONENTIAL SMOOTHING WITH TREND ADJUSTMENT Tt = exponentially smoothed trend in period t

Is used to deal with a systematic trend or seasonal
component.
At = actual demand in period t
 = smoothing constant for the average
β = smoothing constant for the trend

EXPONENTIAL SMOOTHING WITH

EXPONENTIAL SMOOTHING WITH
TREND ADJUSTMENT
TREND ADJUSTMENT
Ft =  (Actual demand of previous period) + (1-  )(Forecast
last period + Trend estimate last period)

Ft=  Dt-1 +(1- )( Ft + Tt) Forecast including trend (FITt ) =

exponentially smoothed forecast (Ft ) + exponentially smoothed

Tt =β(Forecast this period - Forecast last period)+ (1-β)(Trend trend (Tt )
estimate last period)

Tt= β( Ft –Ft-1) +(1-β)(Tt-1)

Period Month Demand Exponentil avg Trend(Tt) Forecast(FIT)

Q3. Table below shows the demand data for a particular model of refrigerator
collected by XYZ Limited. Develop a trend adjusted exponential smoothing T0 = 5
model using  = 0.2 and β=0.3 and determine forecast for next January.
1 Jan 37 37 5 42

Period Month Demand 2 Feb 40 41

1 January 37 3 Mar 41
2 February 40 4 Apr 37
3 March 41
4 April 37
5 May 45 Ffeb=  Dt-1 +(1- )( F0+T0)
6 June 50 Ffeb= 0.2(37) +(1- 0.2)( 37+5)= 41
7 July 43
8 August 47
9 September 56 T1= β( F1 –F0) +(1-β)(T0)
10 October 52
11 November 55 Tfeb= .3( 41 -37) +(1-0.3)(5)=4.7
12 December 54

Forecast FITfeb = F1+T1 =41+4.7=45.7

7
 08-10-2020

Period Month Demand Exponentil avg Trend(Tt) Forecast(FIT) Period Month Demand Exponential avg Trend(Tt) Forecast(FIT)

1 Jan 37 F0 =37 T0 = 5 42 1 Jan 37 F0 =37 T0 = 5 42

2 Feb 40 41 4.7 45.7 2 Feb 40 41 4.7 45.7
3 Mar 41 3 Mar 41 44.56 4.358 48.918
4 Apr 37 4 Apr 37

Fmar=  Dfeb +(1- )( Ffeb+Tfeb) Fapr=  Dmar +(1- )( Fmar+Tmar)
Fmar= 0.2(40) +(1- 0.2)( 41+4.7)= 44.56 Fapr= 0.2(41) +(1- 0.2)( 48.918)= 47.3344

Tmar= β( Fmar –Ffeb) +(1-β)(Tfeb) Tapr= β( Fapr –Fmar) +(1-β)(Tmar)

T1= .3( 44.56 -41) +(1-0.3)(4.7)=4.358 T1= 0.3( 47.3344 -44.56) +(1-0.3)(4.358)=3.88292

Forecast FIT = F1+T1 =44.56+4.358=48.918 Forecast F = Ft+Tt =47.3344+3.88292=51.21732

Period Month Demand Exponential Trend Forecast

average
1 Jan 37 F0 =37 T0 = 5 42 Fjan=  Ddec +(1- )( Fdec+Tdec)
2 Feb 40 41 4.7 45.7 Fjan= 0.2(54) +(1- 0.2)(57.23903)=56.59122
3 Mar 41 44.56 4.358 48.918
4 Apr 37 47.3744 3.388292 51.21732
5 May 45 48.37386 3.02881 51.40374
Tjan= β( Fjan –Fdec) +(1-β)(Tdec)
6 Jun 50 50.12299 2.645657 52.76865 T1= 0.3( 56.59122 -56.08893) +(1-0.3)(1.150092)
7 July 43 52.2149 2.479538 54.69445 = 0.955751
8 Aug 47 52.35556 1.777871 54.1343
9 Sep 56 52.70675 1.349986 54.05661
10 Oct 52 54.4459 1.466468 55.91176
11 Nov 55 55.1294 1.23176 56.36117 Forecast Ft = Ft + Tt
12 Dec 54 56.08893 1.150092 57.23903
=57.546971
56.59122 0.955751