Production and operation

management
Assignment 2

Submitted by:
Saima Jafari
MUM13AP13
BFT 5
NIFT MUMBAI

National Institute of Fashion Technology

Department of Fashion Technology
Production and Operations Management
B. F. Tech V - Assignment II
***
Date of Submission: 03.11.2015
Max. Marks: 20
Caselet - Kalpana Engineering Institute
It was late evening of March 2015, when Dr B K Ramachandran, President of Kalpana
Engineering Institute, a well known engineering institute based in southern India,
had called a meeting with Dr Rajat Bhandari, Registrar and Deans of various
academic departments. There were delays in the construction of new hostel block
and the contractor had shown his inability to hand over the new block before
October 2015. Hostel availability was a major concern for the management to
accommodate students joining the ensuing academic session which was the main
agenda of the meeting.
The management is seeking the best method to forecast students enrolment and take
few buildings on lease outside the campus to accommodate new students. The
enrolment data for the past 05 years was discussed and is given below:
Year
2010
2011
2012
2013
2014

Enrolment
4,500
5,000
5,200
5,600
5,800

Six years ago, the forecast for year 1 was 4,100 enrolments, and the estimated trend
was an increase of 200 per year.
Deliberations included following options:
1. Exponential smoothing forecasting model with = 0.6
2. Exponential smoothing forecasting model with = 0.9
3. Three- year moving average.
4. Three-year weighted moving average, using weights (3/6), (2/6), and (1/6), with
more recent data given more weight.
5. Regression analysis
Questions:
1. Compare above five different approaches. Which method would you recommend
and why?
(1 + 1 + 1 + 1 + 3 + 5)
2. If MAD (mean absolute deviation) is the performance criterion chosen by the
management, which forecasting method should it choose?
2
3. If MSE (mean squared error) is the performance criterion chosen by the
management, which forecasting method should it choose?
2
4. If MAPE (mean absolute percent error) is the performance criterion chosen by
the management, which forecasting method should it choose?
2

SOLUTIONS:
Q.1. Comparison of the 5 different approaches
Exponential smoothing forecasting model with = 0.6
Formula used: F (t+1) = Dt + (1- ) Ft
F 2011 = D2010 + (1- ) F2010
= 0.6 x 4500 + (1- 0.6) x 4100
= 2700 + 1640
= 4340
F 2012 = D2011 + (1- ) F2011
= 0.6 x 5000 + (1- 0.6) x 4340
= 3000 + 1736
= 4736
F 2013 = D2012 + (1- ) F2012
= 0.6 x 5200 + (1- 0.6) x 4736
= 3120 + 1894.4
= 5014.4
F 2014 = D2013 + (1- ) F2013
= 0.6 x 5600 + (1- 0.6) x 5014.4
= 3360 + 2005.76
= 5365.76
F 2015 = D2014 + (1- ) F20145
= 0.6 x 5800 + (1- 0.6) x 5365.76
= 3480 + 2146.304
= 5626.304
5626
Exponential smoothing forecasting model with = 0.9
Formula used: F (t+1) = Dt + (1- ) Ft
F 2011 = D2010 + (1- ) F2010
= 0.9 x 4500 + (1- 0.9) x 4100
= 4050 + 410
= 4460
F 2012 = D2011 + (1- ) F2011
= 0.9 x 5000 + (1- 0.9) x 4460
= 4500 + 446
= 4946
F 2013 = D2012 + (1- ) F2012
= 0.9 x 5200 + (1- 0.9) x 4946

= 4680 + 494.6
= 5174.6
F 2014 = D2013 + (1- ) F2013
= 0.9 x 5600 + (1- 0.9) x 5174.6
= 5040 + 517.46
= 5557.46
F 2015 = D2014 + (1- ) F20145
= 0.9 x 5800 + (1- 0.9) x 5557.46
= 5220 + 555.746
= 5775.746
5776
Three- year moving average
F2013= (D2010 + D2011 + D2012)
3
= (4500+5000+5200)
3
=4900

F2014 = (D2011 + D2012 + D2013)

3
= (5000+5200+5600)
3
=5266.66
5267

F2015 = (D2012 + D2013 + D2014)

3
= (5200+5600+5800)
3
=5533.33
5533
Summarizing,

Year
2010
2011
2012
2013
2014
2015

Enrolment
4,500
5,000
5,200
5,600
5,800
-

3 year Moving Average

Insufficient data
Insufficient data
Insufficient data
4900
5267
5533

Three-year weighted moving average, using weights (3/6), (2/6), and (1/6), with
more recent data given more weight.
F2013= (3/6 * D2012) + (2/6 * D2011) + (1/6 * D2010)
= (3/6 * 5200) + (2/6 * 5000) +(1/6 * 4500)
=5016.66
5017

F2014= (3/6 * D2013) + (2/6 * D2012) + (1/6 * D2011)

= (3/6 * 5600) + (2/6 * 5200) +(1/6 * 5000)
= 5366.66
5367

F2015= (3/6 * D2014) + (2/6 * D2013) + (1/6 * D2012)

= (3/6 * 5800) + (2/6 * 5600) + (1/6 * 5200)
= 5633.33
5633
Summarizing,
Year
2010
2011
2012
2013
2014
2015
Regression analysis

Enrolment
4,500
5,000
5,200
5,600
5,800
-

3 year moving weighted average

Insufficient data
Insufficient data
Insufficient data
5017
5367
5633

Equation of the line of regression: y=a+bt

where,
y= enrolments(demands) in each year
t= time, year
b= change in y with the change in t (slope)
a=y intercept
Formulae used:
y= na+ b t
yt= a t+ b t2
T
1
2
3
4
5
t=15

y
4,500
5,000
5,200
5,600
5,800
y=26,100

yt
4,500
10,000
15,600
22,400
29,000
yt= 81,500

t2
1
4
9
16
25
t2= 55

Putting the values in the equations:

26100 = 5* a + b*15
26100 = 5a+15b
81500 = a*15 + b*55
81500 = 15a+ 55b
On solving these equations, we get:
a= 4260
b= 320
y= 320x + 4260
For 2015, t = 6.
On putting its value:
y = a + bt
y = 4260 + (320 * 6)
= 4260 + 1920
= 6180
Method

Exponential
smoothing
=0.6

Exponential
smoothing
=0.9

Simple
moving
average

Weighted
moving
average

Regression
analysis

Forecast
2015

5626

5776

5533

5633

6180

Recommended method:
The recommended method for the above problem is Regression analysis because of
the following reasons:
Assertion 1: The previous 5 years data and also the estimation that was made 6 years
ago, both say that there is an increasing trend in the number of enrolments (demands)
each year. Only the Regression analysis method gives a forecast of 2015 (i.e. 6180)
which is greater than forecast of 2014 (i.e. 5800) and fulfills the trend pattern.
Assertion 2: If we calculate the forecasts of previous years, we see that regression
analysis method gives a forecast which is closer to the actual demand of that year.
t
1
2
3
4
5

Forecast (y=a+bx)
4260 + (320*1) = 4580
4260 + (320*2) = 4900
4260 + (320*3) = 5220
4260 + (320*4) = 5540
4260 + (320*5) = 5860

Demand
4500
5000
5200
5600
5800

i.e. the errors in this this method are less. It would be more evident from the tables
given below.
Assertion 3: Before choosing the forecasting method, checking the validity of
relationship between variables is important. If we calculate the coefficient of
determination for time period t= 1,2,3,4,5 and demand y
We get, r2= ay + bty- n (y-bar)2
y2 n (y-bar) 2
=

= (4260x26100) + (320x81500) (5x5220 2)

137290000- (5x52202)

0.977, which is a very strong correlation between the increasing time period and
demand, which supports the fact that linear regression analysis should be the preferred
forecasting method.

FORECASTING ERRORS:
Formulae for finding the errors:
(Error) = (Actual demand) (Forecast)
% error = (forecast error / demand) x 100%
TABLE 1: Error table for Exponential smoothing when =0.6
Period
2010
2011
2012
2013
2014

Demand
4500
5000
5200
5600
5800

Forecast
4100
4340
4736
5014.4
5365.76

Error (e)
400
660
464
585.6
434.24
e=2543.74

e2
160000
435600
215296
342927.4
188564.4
2
e = 1342387.8

% error
8.89%
13.2%
8.92%
10.46%
7.49%
e%=48.96%

TABLE 2: Error table for Exponential smoothing when =0.9

Period
2010
2011
2012
2013
2014

Demand
4500
5000
5200
5600
5800

Forecast
4100
4460
4946
5174.6
5557.46

Error (e)
400
540
254
425.4
242.54
e=1861.94

e2
160000
291600
64516
180965.2
58825.65
2
e = 755906.85

% error
8.89%
10.8%
4.88%
7.6%
4.18%
e%=36.35%

TABLE 3: Error table for simple moving average (n=3):

Year Enrolment
2010
4,500
2011
5,000
2012
5,200
2013
5,600
2014
5,800
2015
-

3 year Moving Average

Insufficient data
Insufficient data
Insufficient data
4900
5267
5533

Error e
e2
700
490000
533
284089
e=1233 e2= 774089

% error
12.5%
9.19%
e%=21.69%

TABLE 4: Error table for weighted moving average (n=3):

Year

Enrolment

3 year moving
Weighted average

Error e

e2

% error

2010
2011
2012
2013
2014
2015

4,500
5,000
5,200
5,600
5,800
-

Insufficient data
Insufficient data
Insufficient data
5017
5367
5633

583
433
e=1016

339889
187489
e2= 527378

10.41%
7.47%
e%=17.88%

TABLE 5: Regression analysis:

By putting the values of t=1,2,3,4,5 in the equation y=a+ bt, we can calculate the
forecast for each year. The error table is as follows:
t
1
2
3
4
5

Forecast
4260 + (320*1) = 4580
4260 + (320*2) = 4900
4260 + (320*3) = 5220
4260 + (320*4) = 5540
4260 + (320*5) = 5860

Demand Error (e)

|e|
e2
|Error%|
4500
-80
80
6400
1.78%
5000
100
100
10000
2.00%
5200
-20
20
400
0.38%
5600
60
60
3600
1.07%
5800
-60
60
3600
1.03%
2
e=1016 |e|=320 e =24000 |e%|=6.26%

The values from the above tables are taken for the calculations.
Q.2. If MAD (mean absolute deviation) is the performance criterion chosen by the
management, which forecasting method should it choose?

MAD of Exponential smoothing when =0.6

2543.74 /5= 508.75

MAD of Exponential smoothing when =0.9

1861.94 /5= 372.38

MAD of simple moving average, n=3:

Insufficient data to calculate for required time period

MAD of weighted moving average, n=3:

Insufficient data to calculate for required time period

MAD of Regression analysis:

320 /5= 64
Ans: regression analysis method should be chosen as it gives the minimum mean
absolute deviation

Q.3. If MSE (mean squared error) is the performance criterion chosen by the
management, which forecasting method should it choose?

MSE of Exponential smoothing when =0.6

1342387.8 /5= 268477.55

MSE of Exponential smoothing when =0.9

755906.85 /5 =151181.37

MSE of simple moving average, n=3

Insufficient data to calculate for required time period

MSE of weighted moving average, n=3

Insufficient data to calculate for required time period

MSE of Regression analysis:

24000 /5= 4800

Ans: regression analysis method should be chosen as it gives the minimum mean
squared error
Q.4. If MAPE (mean absolute percent error) is the performance criterion chosen
by the management, which forecasting method should it choose?

MAPE of Exponential smoothing when =0.6

48.96 /5= 9.79

MAPE of Exponential smoothing when =0.9

36.35 /5 = 7.27

MAPE of simple moving average, n=3

Insufficient data to calculate for required time period

MAPE of weighted moving average, n=3

Insufficient data to calculate for required time period

MAPE of Regression analysis:

6.26 /5= 1.25

Ans: regression analysis method should be chosen as it gives the minimum mean
absolute percent error