Professional Documents
Culture Documents
2016
American Journal of Engineering Research (AJER)
e-ISSN: 2320-0847 p-ISSN : 2320-0936
Volume-5, Issue-7, pp-300-311
www.ajer.org
Research Paper
Open Access
Mechanical Engineering Department, Nnamdi Azikiwe University PMB 5025 Awka, Anambra State, Nigeria.
2
Mechanical Engineering Department, Federal Polytechnic Nekede Owerri, Imo State, Nigeria.
ABSTRACT: In this paper, the production quantity of four selected products in British American Tobacco,
Ibadan, Nigeria, was forecasted using the winters forecasting model, in order to foresee the future production
quantities of their products for the next five years (2015-2019). Time series decomposition analysis was used to
study the trend and seasonality in the existing data gotten from the case study company from (2010-2014). The
total result obtained for five years based on the data analysed for the production quantity of each product for
the period of five years were presented in order of actual and winters result respectively. X1=28957095,
X1=30307473, X2=6228361, X2=8465290, X3=9530198, X3=11462915, X4=2041990, X4=-739949. The
results obtained using the adopted model showed that Benson and Hedges(X1), London menthol(X2) and
Rothmans(X3) will be increasing gradually in (their quantity output) for the next five years (2015-2019) while
Pall Mall(X4) will be reducing gradually in their production for the next five years (2015-2019). There is indeed
a significant increase from the results in the production output of the selected products using the winters model
as against the actual production output of the products, as could be seen in the time series decomposition trend
which justifies the research work.
I.
INTRODUCTION
The use of tobacco as generically known is as early as the existence of man [1].With the world
population increasing rapidly over the years, the number of smokers has also been on the increase. This mounted
enormous pressure on the manufacturers to meet this ever-rising demand and to make sure that the consumers
are satisfied with the quality of the products. Quality and quantity of products are examples of process
(production) characteristics and they are both immanent to modern economics where business entities are
operating in rapidly changeable and unstable environment.
Forecasting is defined as the use of historical data to determine the direction of future trends [2].
Forecasting can also be defined as the process of making predictions of the future based on past and present data
and analysis of trends [3].Forecasting is essential for decision making, unless insurance is selected to deal with
the future [4]. Statistical analysis of time series data started a long time ago [5], and forecasting has an even
longer history. Objectives of the two studies may differ in some situations but forecasting is often the goal of a
time series analysis. A common place example might be estimation of some variable of interest at some
specified future date. Prediction is a similar but more general term. Both might refer to formal statistical
methods or longitudinal data or alternatively to less formal judgmental methods. Usage can differ between areas
of application: for example, in hydrology, the teams forecast and forecasting are sometimes reserved for
estimates of values at certain specific future times. While the term prediction is used for more general
estimates, such as the number of times flood will occur over a long period [6].
Forecasting is divided into two categories, which are qualitative and quantitative methods [7].
Qualitative methods are subjective, based on the opinion and judgment of consumers and experts. They are
appropriate when past data are not available. Quantitative forecasting methods are used to forecast future data as
a function of past data. They are appropriately used when past numerical data is available and when it is
reasonable to assume that some patterns are expected to continue into the future. These methods are usually
applied to start or inter mediate range decisions. Examples of quantitative forecasting models are last period
demand, simple exponential smoothing and multiplicative seasonal indexes etc. Forecasting models can be
broken into approaches like average approach, casual econometric methods, nave approach, drift method,
seasonal nave approach, artificial intelligence methods and time series methods [8].
www.ajer.org
Page 300
2016
Time series methods use historical data as the basis of estimating future outcomes. Examples are:
Moving average, Weighted moving average, Winters model, Exponential smoothing model, Kalmau filtering,
Auto regressive moving average (ARMA), Autoregressive integrated moving average (ARIMA), Extrapolation,
Linear prediction, Trend estimation, Growth curve (statistics). A time series is a sequence of data points,
typically consisting of successive measurements made over a time interval [9].Time series are used in
mechanical engineering, statistics, signal processing, pattern recognition, econometrics, mathematical finance,
weather forecasting, intelligent transport and trajectory forecasting, earthquake prediction, electro
encephalography, control engineering astronomy, communications engineering and largely in any domain of
applied science and engineering [9].
Exponential Smoothing is a procedure for continually revising a forecast with light of none recent
experience [8].It assigns exponentially decreasing weights as the observation get older. In other words, recent
observations are given relatively no weight in forecasting than the older observations. This is a rule of thumb
technique for smoothing time series data, particularly for computing in a finite amount of time by applying as
many as three low-pass filters (three filters that passes signals with a frequency lower than a certain cut off
frequency and reduces in degree the signals with frequencies higher than the cut off frequency) with exponential
window functions [10]. Exponential Smoothening is categorized into three, which are; single, double and triple
exponential smoothening. The single exponential smoothening is used for short range forecasting, usually first
one month into the future, whereas the Double Exponential Smoothening is used when the data shows a trend.
For data showing trend and seasonality, the Triple Exponential Smoothening, also called the Winters Model is
used [11].
A few examples of works on forecasting includes the Sales forecast in Apparel and Fashion industry
using data mining techniques, extrapolation techniques based on pre-sales and triple exponential smoothing [12]
and forecasting of the production quantity done on Louis Carter plastic industry using historical data [13].
Having seen so many researches in the area of forecasting especially in industries, this paper is therefore aimed
at the forecasting of production quantity of some selected tobacco products in a tobacco industry using the
double exponential smoothing model and winters model as the forecasting techniques/ models.
II.
METHODOLOGY
www.ajer.org
Page 301
2013
2014
Jan
Feb
Mar
April
May
June
July
Aug
Sept
Oct
Nov
Dec
Jan
Feb
Mar
April
May
June
July
Aug
Sept
Oct
Nov
Dec
Jan
Feb
Mar
April
May
June
July
Aug
Sept
Oct
Nov
Dec
Total
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
481,513
493,674
511,846
581,633
590,244
514,382
594,445
400,617
451,783
478,271
400,667
318,799
493,460
593,611
586,423
553,007
596,178
534,629
680,442
492,337
500,681
511,556
397,442
300,651
449,650
451,880
555,674
434,658
574,416
503,217
684,346
481,229
500,391
511,372
332,116
294,226
28957095
59,834
133,636
130,951
110,107
107,174
122,546
59,926
142,000
50,890
74,317
107,936
92,000
120,094
142,671
134,562
80,626
78,712
120,167
109,169
126,445
124,339
139,452
119,697
87,457
140,980
52,130
124,780
73,860
109,510
122,671
110,231
131,442
126,317
130,119
117,143
90,116
6228361
92,681
174,976
184,293
244,164
160,546
249,823
154,126
114,173
124,096
202,630
84,900
181,317
138,469
300,116
211,793
146,555
312,446
200,662
186,472
100,652
176,442
198,279
146,476
148,568
106,340
162,210
166,948
135,565
157,300
106,360
85,490
142,660
156,379
172,600
144,372
143,006
9530198
2016
0
39986
35,374
32,297
42,189
23,174
54,570
0
0
45,026
16,111
11,780
47,223
42,453
34,230
61,169
46,432
38,631
44,662
27,851
36,781
54,972
31,930
30,881
25,450
33,070
37,260
28,712
26,100
30,448
46,323
45,118
62,390
27,114
30,020
0
2041990
www.ajer.org
Page 302
2016
III.
RESULTS ANALYSIS
V a r ia b le
A c tu a l
F its
600000
T ren d
A c c u r a c y M e a su r e s
MA PE
X 1 ( s t ic ks )
500000
MA D
38230
MSD
2536357839
400000
300000
200000
100000
1
12
18
24
30
36
42
48
54
60
In d e x
www.ajer.org
Page 303
2016
trend which is the green spotted on the figure is seen to be increasing gradually and this is a sign that the actual
future production of the X1 sticks is on an increasing side.
Table 3: Time Series Decomposition for X2(sticks)
Data
X2(sticks)
Accuracy Measures
Length 60
MAPE
18
NMissing 0
MAD
14604
MSD 517174737
Seasonal Indices
Period Index
1 1.11101
2 1.26234
3 1.21475
4 0.68719
5 0.80723
6 1.04551
7 0.92451
8 1.08452
9 1.08395
10 0.98928
11 1.01651
12 0.77320
From The Time series decomposition for X2 sticks as in table 3 above, it can be deduced that the 4th
and 2 months have the lowest and highest seasonal index of 0.68719 and 1.26234 respectively. This implies
that the 4th month is not affected much by season which gives a better production output as compared to the 2 nd
month which has the highest seasonal index and consequently negatively affected production output.
nd
T im e S e r ie s D e c o m p o s itio n P lo t f o r X 2 ( s tic k s )
M u ltip lic a tiv e M o d e l
V a r ia b le
140000
A c tu a l
F its
130000
T ren d
120000
A c c u r a c y M e a su r e s
X 2 ( s t ic ks )
MA PE
110000
18
MA D
14604
MSD
517174737
100000
90000
80000
70000
60000
50000
1
12
18
24
30
36
42
48
54
60
In d e x
www.ajer.org
Page 304
2016
5 1.37024
6 1.12466
7 0.96117
8 0.56126
9 1.02646
10 1.15282
11 0.87740
12 0.91993
In the Time Series Decomposition for X3(sticks) in table 4 above, the month with the lowest seasonal
index (0.56126) is the 8th month, which implies that it is not affected much by season which gives a better
production output when compared to the 2ndmonth with the highest seasonal index of 1.23125 which will have
its production output affected negatively due to the effect of season.
T im e S e r ie s D e c o m p o s itio n P lo t f o r X 3 ( s tic k s )
M u l ti p l i c a t i v e M o d e l
V a r ia b le
A c tu a l
300000
F its
T ren d
A c c u r a c y M e a su r e s
X 3 ( s t ic ks )
250000
MA PE
17
MA D
25489
MSD
1328455961
200000
150000
100000
12
18
24
30
36
42
48
54
60
In d e x
www.ajer.org
Page 305
2016
The time series decomposition for X4 sticks in table 5 above shows that The 12 th month has the lowest
seasonal index (0.85838), which implies that it will not be affected much by season, which gives a better
production output as compared to the 7th month with the highest seasonal index of 1.29320.
T im e S e r ie s D e c o m p o s itio n P lo t f o r X 4 ( s tic k s )
M u ltip lic a tiv e M o d e l
70000
V a r ia b le
A c tu a l
F its
60000
T ren d
A c c u r a c y M e a su r e s
50000
X 4 ( s t ic ks )
M A PE
40000
56
M A D
10936
M SD
202082031
30000
20000
10000
0
1
12
18
24
30
36
42
48
54
60
In d e x
Table 6 above, shows the time which is the length of period i.e. 1-60months, the X1 sticks which is the
actual data of the X1 sticks, the smooth which is the smoothened data of the actual data, the predict which is the
predicted data of the model and the error which is the actual data minus the predicted data.
www.ajer.org
Page 306
2016
W in te r s ' M e th o d P lo t fo r X 1 ( s tic k s )
M u ltip lic a tiv e M e th o d
V a r ia b le
1000000
A c tu a l
F its
F o r e c a s ts
800000
95.0%
PI
X 1 ( s t ic ks )
S m o o th in g C o n s ta n ts
600000
400000
A lp h a ( le v e l)
0.2
G a m m a ( tr e n d )
0.2
D e lta ( s e a s o n a l)
0.2
A c c u rac y
M A PE
200000
M e a su r e s
11
M A D
49098
M SD
3811379937
12
24
36
48
60
72
84
96
108
120
In d e x
www.ajer.org
Accuracy Measures
MAPE
22
MAD
18866
MSD 629041633
Page 307
2016
Table 7, it shows the time which is the length of period i.e. 1-60months, the X2 sticks which is the
actual data of the X2 sticks, the smooth which is the smoothened data of the actual data, the predict which is the
predicted data of the model and the error which is the actual data minus the predicted data.
W in te r s ' M e th o d P lo t fo r X 2 ( s tic k s )
M u ltip lic a tiv e M e th o d
350000
V a r ia b le
A c tu a l
300000
F its
F o r e c a s ts
95.0%
250000
PI
X 2 ( s t ic ks )
S m o o th in g C o n s ta n ts
200000
150000
A lp h a ( le v e l)
0.2
G a m m a ( tr e n d )
0.2
D e lta ( s e a s o n a l)
0.2
A c c u r a c y M e a su r e s
M A PE
100000
22
M A D
18866
M SD
629041633
50000
0
12
24
36
48
60
72
84
96
108
120
In d e x
Accuracy Measures
MAPE
19
MAD
28747
MSD 1398896094
Table 8, it shows the time which is the length of period i.e. 1-60months, the X3sticks which is the
actual data of the X3 sticks, the smooth which is the smoothened data of the actual data, the predict which is the
predicted data of the model and the error which is the actual data minus the predicted data.
www.ajer.org
Page 308
2016
W in te r s ' M e th o d P lo t fo r X 3 ( s tic k s )
M u ltip lic a tiv e M e th o d
V a r ia b le
500000
A c tu a l
F its
F o r e c a s ts
400000
95.0%
PI
X 3 ( s t ic ks )
S m o o th in g C o n s ta n ts
300000
200000
A lp h a ( le v e l)
0.2
G a m m a ( tr e n d )
0.2
D e lta ( s e a s o n a l)
0.2
A c c u rac y
M A PE
100000
M e a su r e s
19
M A D
28747
M SD
1398896094
12
24
36
48
60
72
84
96
108
120
In d e x
Accuracy Measures
MAPE
71
MAD
13169
MSD 277357360
Table 9 shows the time which is the length of period i.e. 1-60months, the X4sticks which is the actual
data of the X4 sticks, the smooth which is the smoothened data of the actual data, the predict which is the
predicted data of the model and the error which is the actual data minus the predicted data.
www.ajer.org
Page 309
2016
W in te r s ' M e th o d P lo t fo r X 4 ( s tic k s )
M u ltip lic a tiv e M e th o d
100000
V a r ia b le
A c tu a l
F its
50000
F o r e c a s ts
95.0%
PI
X 4 ( s t ic ks )
S m o o th in g C o n s ta n ts
-50000
A lp h a ( le v e l)
0.2
G a m m a ( tr e n d )
0.2
D e lta ( s e a s o n a l)
0.2
A c c u r a c y M e a su r e s
MA PE
-100000
71
MA D
13169
MSD
277357360
-150000
12
24
36
48
60
72
84
96
108
120
In d e x
2016
2017
Month
Jan
Feb
Mar
April
May
June
July
Aug
Sept
Oct
Nov
Dec
Jan
Feb
Mar
April
May
June
July
Aug
Sept
Oct
Nov
Dec
Jan
Feb
Mar
April
May
June
July
Aug
www.ajer.org
M. unit
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
X1(sticks)
472799
520700
547720
521064
581747
524935
626391
458463
492121
492793
374428
282813
479476
528045
555436
528397
589924
532304
635174
464885
499006
499679
379654
286755
486152
535389
563153
535729
598100
539674
643958
471306
X2(sticks)
114235
120633
142933
100288
103405
132476
114926
144376
122927
129870
126847
99486
122129
128921
152697
107100
110390
141374
122602
153966
131047
138402
135134
105951
130022
137209
162461
113912
117374
150272
130278
163556
X3(sticks)
105166
187758
172456
150700
203053
175033
139883
112648
160079
183682
134726
153606
117347
209297
192052
167663
225697
194372
155197
124869
177290
203256
148956
169689
129527
230835
211648
184626
248341
213712
170512
137091
X4(sticks)
26020.2
31981.0
26070.7
28837.2
26837.7
22505.8
28899.6
16253.5
19200.9
16616.2
12472.2
5437.4
10748.1
12244.9
9109.7
9001.0
7254.3
5020.1
4892.0
1747.1
686.9
-805.5
-1856.5
-1470.8
-4524.1
-7491.2
-7851.2
-10835.2
-12329.0
-12465.5
-19115.7
-12759.4
Page 310
2018
2019
Sept
Oct
Nov
Dec
Jan
Feb
Mar
April
May
June
July
Aug
Sept
Oct
Nov
Dec
Jan
Feb
Mar
April
May
June
July
Aug
Sept
Oct
Nov
Dec
Total
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
505891
506565
384880
290698
492829
542734
570869
543061
606277
547043
652741
477727
512775
513451
390106
294641
499506
550078
578585
550393
614454
554413
661525
484148
519660
520337
395332
298584
30307473
139167
146933
143422
112416
137916
145497
172225
120724
124358
159169
137954
173145
147287
155465
151710
118881
145809
153785
181989
127536
131342
168067
145630
182735
155407
163997
159997
125346
8465290
194502
222830
163187
185773
141708
252374
231244
201589
270985
233051
185826
149312
211713
242404
177418
201856
153888
273912
250840
218553
293628
252390
201141
161534
228925
261978
191648
217939
11462915
2016
-17827.2
-18227.1
-16185.3
-8379.0
-19796.3
-27227.2
-24812.2
-30671.3
-31912.4
-29951.1
-43123.3
-27265.9
-36341.2
-35648.7
-30514.0
-15287.2
-35068.5
-46963.3
-41773.2
-50507.5
-51495.8
-47436.8
-67131.0
-41772.3
-54855.3
-53070.3
-44842.8
-22195.4
-739949
Table 10above shows the monthly forecasting results of all the cigarette brands using the winters
method over a period of five years (2015-2019) as earlier discussed.
IV.
CONCLUSION
The total result obtained using the data for the production quantity of each of the products for the
period of five years in thousands of sticks is presented in order of actual and Winters result respectively,
X1=28957095, X1=30307473, X2=6228361, X2=8465290, X3=9530198, X3=11462915, X4=2041990, X4=739949. From the forecast of the production quantity of four different cigarette brands (Benson and Hedges,
London Menthol, Rothmans and Pall Mall) using Winters method, it can be seen that the future production of
Pall Mall will gradually decrease while the rest will gradually keep on increasing, as could been seen in the time
series decomposition trend.
REFERENCES
[1].
[2].
[3].
[4].
[5].
[6].
[7].
[8].
[9].
[10].
[11].
[12].
[13].
[14].
Ali O. Yusuf., (2001). Tobacco usage and Nigerian laws, Nigerian Bar
Journal, vol. 8, pp. 1-2.
Reekie, H., (1997). Sales Forecasting Systems in Corporate America. Journal of Business Forecasting Methods & System, vol.16,
Iss.1, pp.6-12.
Reinhardt, J.W., & Giles, B., (2001). Smoothing, Forecasting and Prediction, Englewood Cliffs, Pearson Hall, pp 33-74.
Armstrong, J.S., (1999). Forecasting Methods for Marketing: Review of empirical research, International Journal of Forecasting,
vol.3, pp. 355-376.
Yule, E., (1927). Advances in time series forecasting. London, Bentham Science Publishers,pp 5-7.
Cameron, B.C., (1998). Principles of forecasting a handbook for researchers and practitioners. Nurwell Kluwer Academic
Publishers, pp.56-64.
Vollmann, J., (2011). Principles of Forecasting. Hingham, USA, Kluwer Academic Publishers, pp 2-5.
Kahn K. B., (2002). An explanatory investigation of new product forecasting techniques, Journal of Product Innovation
Management, vol.19, Iss.2, pp.133-143.
Paul Karapanagiotidis, (2012). Literature Review of Modern Times Series Forecasting Methods, International Journal of
Forecasting, vol.27, Iss.4, pp.11-79.
Kahn, K.B., (2010). The hard and soft sides of new product forecasting. Journal of Business Forecasting, vol.28, Iss.1, pp.29-31.
Krajewski, et al (2007). "Operations Management : Processes and Value Chain 8thed. Pearson Prentice Hall, pp. 10-12.
SebastienThomassey, (2007). Sales Forecasting in Apparel and Fashion Industry. International Journal of Forecasting, vol.28,
Iss.4, pp.830-841.
Okolie, P.C et al (2015). Analysis and Forecasting of the Production Quantity in a Manufacturing Industry Using Historical Data:
A case study of Louis Carter Plastic Industry. The International Journal of Engineering and Applied Science, vol.4, Iss.10, pp.0717.
PrajaktaKalekar, S., (2004). Time series Forecasting using Holt Winters Exponential Smoothing, Ph.D. research thesis,
KanwalRekhi School of Information Technology, pp.2-8
www.ajer.org
Page 311