Research Article

International Journal of Engineering

Business Management
Volume 10: 1–9
Forecasting of demand using ª The Author(s) 2018
DOI: 10.1177/1847979018808673
ARIMA model journals.sagepub.com/home/enb

Jamal Fattah1, Latifa Ezzine1, Zineb Aman2 ,

Haj El Moussami2, and Abdeslam Lachhab1

Abstract
The work presented in this article constitutes a contribution to modeling and forecasting the demand in a food company,
by using time series approach. Our work demonstrates how the historical demand data could be utilized to forecast future
demand and how these forecasts affect the supply chain. The historical demand information was used to develop several
autoregressive integrated moving average (ARIMA) models by using Box–Jenkins time series procedure and the adequate
model was selected according to four performance criteria: Akaike criterion, Schwarz Bayesian criterion, maximum
likelihood, and standard error. The selected model corresponded to the ARIMA (1, 0, 1) and it was validated by another
historical demand information under the same conditions. The results obtained prove that the model could be utilized to
model and forecast the future demand in this food manufacturing. These results will provide to managers of this man-
ufacturing reliable guidelines in making decisions.

Keywords
Demand forecasting, time series, autoregressive integrated moving average (ARIMA)

Date received: 7 June 2018; accepted: 25 September 2018

Introduction Generally, there are many approaches to forecast

demand among which we find the exponential smoothing,
In today’s competitive manufacturing environment, and to
for example. But, when applying these approaches, we
respond quickly to shifting demand, organizations are mov-
need to have historical data. In the beginning, there is no
ing toward a more effective demand-driven supply chain.
information about the past, we use then an estimation based
The market has evolved into a “pull” environment with
on similar cases or engineer’s experiences. In this case, we
customers more demanding and discriminating, dictating
have a big amount of uncertainty which will be avoided
to the supplier what products they desire and when they
with time.
need them delivered.1
Demand forecasting is crucial to inventory manage-
ment. Inventory stock levels depend on demand’s forecasts.
In fact, inaccurate estimation of demand can cause signif- 1
Modeling, Control Systems and Telecommunications Team, Faculty of
icant costs to pay, which proves that the process is not
Sciences, Moulay Ismail University, Meknes, Morocco
improved. Consequently, many systems incur large invest- 2
Mechanics and Integrated Engineering Team, ENSAM School, Moulay
ments in inventories to avoid “stock outs.” A further com- Ismail University, Meknes, Morocco
plicating issue is that some demands can be intermittent
demands, which means that there is a time when we have Corresponding author:
Zineb Aman, Mechanics and Integrated Engineering Team, ENSAM School,
no demand and other time when we have successive Moulay Ismail University, P.O. Box 15290, Al Mansour, Meknes 50000,
demands. Intermittent demands present many difficulties Morocco.
for traditional statistical demand forecasting methods. Email: [email protected]

Creative Commons CC BY: This article is distributed under the terms of the Creative Commons Attribution 4.0 License
(http://www.creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of the work without
further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/
open-access-at-sage).
2 International Journal of Engineering Business Management

For most organizations, managing demand is challen- material and inventory management. In our case, we will
ging because of the difficulty in forecasting future con- work in a food company—which requires more than in
sumer needs accurately.1 More than 74 % of the responds other sectors—forecasts that are very reliable and accurate
in a research survey, shows the poor forecasting accuracy as long as it will affect the production of perishable goods.
and demand volatility as the increasing major challenges to Besides, products in a food company having steady pre-
supply chain flexibility.2 Best performing companies tend dictable demand need efficient supply chains that shorten
to improve supply chain flexibility, agility, and responsive- lead times and support limited inventory.
ness through improving forecasting accuracy throughout A robust supply chain management system requires the
the long supply chain.2 The managers in these companies presence of managers who are aware of the necessity of
must link forecasting to improvement goals and use past collaboration between different functions: planning, pro-
performance to avoid past errors and then reach a high level curement, manufacturing, and logistics. Let’s take the
of efficiency.3 example of the collaboration of planning function with
Researchers came out with much work in the forecasting suppliers. Over time, we feed our database to build a his-
domain and suggested many methods among which we find tory that will be used to make our forecasts. After devel-
two principal approaches much utilized: time series oping our model, which is the purpose of our article, we
approaches and artificial neural network (ANN) techniques. will easily have the planned application transmitted to the
ANN models have been successfully involved in fore- planning function. The latter carries out its production plan
casting demand. These models are characterized by inter- related to suppliers.
vals with considerable variation of demand. ANN approach The aim is therefore to devise an optimal production
is considered as an alternative when it comes to the ability plan based on accurate forecasts to minimize the total pro-
to capture the nonlinearity in data set. duction cost composed by the procurement, processing,
ANN is applied in different fields. Gaafar and Choueiki4 storage, and distribution costs. Expected benefits from
applied a neural network model to a lot-sizing problem as a these forecasts are reduced inventories, lower supply chain
part of material requirements planning for the case of deter- costs, increased return on assets, greater customer satisfac-
ministic time-varying demand.5 tion, and reduced lead times. However, this optimal pro-
To compare ANN and ARIMA method and to assess the duction plan should meet different company constraints
performance of the two methods, a study related to elec- among others: production capacity, minimum production
tricity demand has been done by Prybutok et al.6 to forecast lots, and so on.
a time series. ANN seems to be outperformed. Another We will be interested in the evolution by making fore-
study was done by Ho et al.7 using simulated failure time casts of the demand in a Moroccan food company. To
of a compressor to determine the more accurate forecasting achieve its objectives, the company must rely on precise
model. The two methods are used to forecast the failure of forecasts. In this context, our article aims mainly to study
the system.8 the demand to provide precise forecasts and to respect the
Aburto and Weber9 combined the two forecasting meth- permissible error margin. The main idea is that forecasting
ods which are ARIMA and neural networks. The efficiency accuracy drives the performance of inventory management.
of the hybrid model is compared with traditional forecast- The aim of the present study is the modeling and fore-
ing methods.10 casting of demand by using Box–Jenkins time series
This brief review of the literature shows that ANN is a approach, especially the ARIMA. To achieve this goal,
strength tool aiming at the modeling of any time series. we used large and consistent historical demand data: from
Nevertheless, in our article, we will test the ARIMA model January 2010 until December 2015. Several ARIMA mod-
at first to prove its ability to make accurate forecasts in the els were developed and evaluated by four performance
food company as a priori study. criteria: Akaike criterion (AIC), Schwarz Bayesian criter-
In our article, we are interested the most in the time ion (SBC), maximum likelihood, and standard error. The
series approach: autoregressive integrated moving average adequate model was validated by new historical demand
(ARIMA) models,11–14 multivariate transfer function mod- data under the same conditions. In this article, the second
els,11,15 dynamic models,11 and generalized autoregressive section presents a literature review about demand forecast-
conditional heteroskedasticity (GARCH) models16 have ing studies. The third section is consecrated to the results
also been proposed. Certainly, ARCH and GARCH models and discussions of our case study. Finally, the article con-
are increasingly utilized and are considered as important cludes with a summary and the future work.
tools in the analysis of time series data, especially in the
case of financial applications. But, they are specifically
dedicated to the analysis and forecasting of volatility which Literature review
is not our aim in this current article.
In all sectors, demand forecasts are of great importance.
Forecasting demand
Indeed, predicting the demand facilitates the decision on In today’s organizations, which are subject to abrupt and
the amount to produce and thus on the supply of the raw enormous changes that affect even the most established of
Fattah et al. 3

structures and where all requirements of business sector model to improve forecasting accuracy, the seasonal fac-
need accurate and practical reading into future, the fore- tors are calculated from multiplicative model. Hyndman33
casts are becoming very crucial since they are the sign of widened Miller and Williams’32 work by applying different
survival and the language of business in the world. A fore- relationships between trend and seasonality under seasonal
cast is a science of estimating the future level of some ARIMA hypothesis. The classical ARIMA approach
variables. The variable is most often demand, but it can becomes prohibitive, and in many cases, it is impossible
also be something else, such as supply or price.17 Forecast- to determine a model, when seasonal adjustment order is
ing is the operation of making assumption about the future high or its diagnostics fail to indicate that time series is
values of studied variables.18 stationary after seasonal adjustment. In such cases, the
In manufacturing, forecasting demands is among the static parameters of the classical ARIMA model are
most crucial issues in inventory management19; it can be considered the principal constraint to forecasting high
used in various operational planning activities during the variable seasonal demand. Another constraint of the
production process: capacity planning, used-product acqui- classical ARIMA approach is that it requires a large
sition management.20 number of observations to determine the best fit model
For both types of supply chain processes “push/pull,” the for a data series.
demand forecasts are considered the ground of supply An ARIMA model is labeled as an ARIMA model (p, d,
chain’s planning. The pull processes in the supply chain are q), wherein:
realized with reference to customer demand, while all push
processes are realized in anticipation of customer demand.21  p is the number of autoregressive terms;
A company must take into consideration such factors before  d is the number of differences; and
selecting a suitable forecasting methodology because the  q is the number of moving averages.
choice of a methodology is not as simple as it seems. Fore-
casting methods are categorized according to four types: The autoregressive process. Autoregressive models assume
qualitative, time series, causal, and simulation.21 that Yt is a linear function of the preceding values and is
A time series is nothing but observations according to the given by equation (1)
chronological order of time.17 Time series forecasting mod-
els use mathematical techniques that are based on historical Yt ¼ α 1 Yt1 þ εt ð1Þ
data to forecast demand. It is founded on the hypothesis that Literally, each observation consists of a random compo-
the future is an expansion of the past; that’s why we can nent (random shock, ε) and a linear combination of the
definitely use historical data to forecast future demand.1 previous observations. α 1 in this equation is the self-
Many studies about demand forecasting by time series regression coefficient.
analysis have been done in several domains. They encircle
demand forecasting for food product sales,22 tourism,23 The integrated process. The behavior of the time series
maintenance repair parts,19,24 electricity,25,26 automobile,27 may be affected by the cumulative effect of some pro-
and some other products and services.28,29,30 cesses. For example, stock status is constantly modified
By time series analysis, the forecasting accuracies depend by consumption and supply, but the average level of stocks
on the characteristics of time series of demand. If the transi- is essentially dependent on the cumulative effect of the
tion curves show stability and periodicity, we will reach high instantaneous changes over the period between inventories.
forecasting accuracies, whereas we can’t expect high accura- Although short-term stock values may fluctuate with large
cies if the curves contain highly irregular patterns.27 contingencies around this average value, the level of the
series over the long term will remain unchanged. A time
series determined by the cumulative effect of an activity
Autoregressive integrated moving average
belongs to the class of integrated processes. Even if the
To model time series, we can work with the traditional sta- behavior of a series is erratic, the differences from one
tistical models including moving average, exponential observation to the next can be relatively low or even oscil-
smoothing, and ARIMA. These models are linear since the late around a constant value for a process observed at dif-
future values are cramped to be linear functions of past data. ferent time intervals. This stationarity of the series of
During the past few decades, researchers have been differences for an integrated process is a crucial character-
focusing much on linear models since they had proved istic viewed from the statistical analysis side of the time
simplicity in comprehension and application. series. Integrated processes are the archetype of nonstation-
Time series forecasting models are mostly used to pre- ary series. A differentiation of order 1 assumes that the
dict demand. Under an autoregressive moving average difference between two successive values of Y is constant.
hypothesis, Kurawarwala and Matsuo31 calculated the sea- An integrated process is defined by equation (2)
sonal variation of demand by using historical data and vali-
Yt ¼ Yt1 þ εt ð2Þ
dated the models by examining the forecast performance.
Miller and Williams32 mixed seasonal factors in their where the random perturbation εt is a white noise.
4 International Journal of Engineering Business Management

The moving average process. The current value of a moving

averaging process is a linear combination of the current
disturbance with one or more previous perturbations. The
moving average order indicates the number of previous
periods embedded in the current value. Thus, a moving
average is defined by equation (3)
Yt ¼ εt  θ 1 εt1 ð3Þ
34
Box and Jenkins was founded on the contributions of
Yule35 and Wold36 to develop a practical approach in order
to perform ARIMA models. The Box–Jenkins principle
consists of three iterative steps of model identification,
parameter estimation, and diagnostic checking steps37. The
principle rule to identify the model is that if a time series is
obtained from an ARIMA process, it should have some
theoretical autocorrelation properties. By matching the the-
oretical and empirical autocorrelation patterns, we make it
possible to identify one or several potential models for the Figure 1. Evolution of the final product’s sales.
given time series. Box and Jenkins34 proposed to use the
autocorrelation function (ACF) and the partial autocorrela- and the accuracy and characteristics are studied. This study
tion function (PACF) of the sample data as the basic tools examines the effectiveness of demand forecasting in a food
to identify the order of the ARIMA model. manufacturing.
As far as the identification step is concerned, we should Based on the Box–Jenkins approach, our study will be
produce a stationary time series, which is a required con- carried out in three parts: identification, estimation, and
dition to find the ARIMA model, so, we mostly need data verification. The model shown in Figure 1 is based on the
transformation. The statistical characteristics of a station- demand of the final product in a Moroccan food manufac-
ary time series such as the mean and the autocorrelation turing from January 2010 until December 2015.
structure are constant over time. We usually need to apply
differencing and power transformation to the data to
remove the trend and stabilize the variance before an Identification of model
ARIMA model can be fitted. In this step, we start with the initial preprocessing of the
After that, it becomes easy to calculate the model para- data to make it stationary, and then we choose possible
meters and then specify the model. These parameters are values of p and q which we can of course adjust as model
estimated so that the overall error is reduced. fitting progresses.
Finally, we move to the diagnostic checking of model For stationarity, the series shown in Figures 2 and 3,
adequacy. In this last step, we make sure that the hypothesis respectively, fluctuates around an average value and its
we made about the errors are contended. The diagnostic ACF decays to zero fairly rapidly which proves the statio-
statistics and plots of residuals can be used to assess the narity of the time series.
adequacy of future values to our data. If the model is not Moreover, to assess whether the data come from a sta-
adequate, we have to make other estimation of parameters tionary process we can perform the unit root test: Dickey–
followed by the model validation. Diagnostic information Fuller test for stationarity. After carrying out the test on the
can help us come up with new models. Xlstat software, the results are grouped in Table 1.
Box–Jenkins model constitutes a process approach to
follow and repeat till reaching a high degree of satisfaction H0: The series has a unit root.
about the model and having reduced errors. Consequently, H1: The series does not have a unit root. The series is
we can freely use this model to forecast our variable. stationary.
Researchers approve that the estimation of parameters
requires a large number of observations. Consequently, Since the calculated p value is greater than the threshold
there are some limits for using ARIMA model. Neverthe- significance level α ¼ 0.05, the null hypothesis H0 cannot
less, once we apply ARIMA model, we reach a high quality be rejected. The risk of rejecting the null hypothesis H0
in the opposite of the time series models. while it is true is 84.38%.
In our case, and after verification of the stationarity of
the series, we notice from the ACF and PACF correlo-
Results and discussion grams that our model is not pure AR or pure MA. We
In this article, the demand forecasting of the final product therefore tested several models to identify the most suit-
in a food manufacturing is conducted based on real data, able one for sales.
Fattah et al. 5

intervals of the adjustment at a 95% confidence level.

The best model is as simple as possible and minimizes
certain criteria, namely AIC, SBC, variance and maxi-
mum likelihood. 43–45 The chosen model is that of
ARIMA (0, 1, 1). For the other models, either Student
“T-RATIO” test values are found in the range +1.96, or
one of the values of the minimization criteria is higher
than that found for the ARIMA model (1, 0, 1) with the
constant value.
Table 2 summarizes the values of the different mod-
els and proves the choice of the model on which we will
base our predictions. It is clear from Table 2 that the
ARIMA model (1, 0, 1) is selected because all the coef-
ficients are significantly different from 0 according to
the Student test (|T-RATIO|  1, 96) with an acceptable
level of adjustment.
The model residue is stationary and follows a white
Figure 2. ACF correlograms of the demand series. noise process in the range of +40. The residue histogram
ACF: autocorrelation function. shows whether the distribution of residues approximates a
normal distribution. In our case, we have residues that dis-
tribute relatively normal around zero and with a relatively
low dispersion at a 5% risk.
The chosen model parameters are presented in Table 3.
The developed model is given by equation (4)
yt ¼ δ þ α 1 yt1  θ 1 εt1 þ εt ð4Þ
with:

 yt, yt1 : sales of period t and t1, respectively.

 εt ; εt1 : residuals of period t and t1, and constitute
a white nose.
 α 1 , θ 1 : coefficients of autoregressive and moving
average processes, respectively.

From Table 3, we can extract the coefficients of auto-

regressive and moving average processes. Therefore, equa-
tion (4) becomes as follows
Figure 3. PACF correlograms of the demand series. yt ¼ 125:524 þ 0:90792 yt1  0:6388 εt1 ð5Þ
PACF: partial autocorrelation function.

Table 1. Test results.

τ (observed value) 1.350

Accuracy of ARIMA (1, 0, 1) model
τ (critical value) 0.717 The accuracy of the developed model was evaluated by
p Value (unilateral) 0.844 comparing the experimental and the simulated sales in the
α 0.05 same period. Figure 4 reports this comparison and reveals
that the selected model has a high accuracy and ability to
simulate the dynamic behavior of sales. Therefore, this
Estimation of model’s coefficients model can be used to analyze and model the demand in
The ARIMA procedure of the SPSS time series module38 this food manufacturing.
allows estimating the coefficients of the models that we From the graph, we notice that the model is validated
have previously identified by providing the parameters p, since the predicted demand fluctuates around the fit. We
q, and d, using a fast maximum likelihood estimation also note that the predicted demand stayed between the
algorithm.39–42 upper limit and the lower limit.
The execution of the procedure adds new time series We can see that the error variates but it is among the
representing the values adjusted or predicted by the tolerance interval. In order to minimize this error, we pro-
model, residuals (adjustment errors) and confidence pose a new approach in our future work.
6 International Journal of Engineering Business Management

Table 2. Coefficients of different models.

Models ARIMA (1, 0, 1)

without
Characteristics ARIMA (1, 0, 2) ARIMA (2, 0, 2) ARIMA (1, 0, 1) ARIMA (1, 0, 0) ARIMA (0, 0, 1) constant
AR (1) α1 0.92913 0.71371 0.90792 0.49434 0.41704 0.99755623
SEB 0.104616 0.761758 0.094852 0.1074471 0.1119989 0.00444769
T value 8.8813204 0.9369292 9.571955 4.600820 3.723567 224.28617
p Value 0.00000000 0.35216008 0.00000000 0.00001823 0.00039384 0.0000000
MA (1) θ1 0.52269 0.31779 0.63880 0.71392452
SEB 0.167073 0.741595 0.161531 0.08579173
T value 3.1284995 0.4285186 3.954655 8.32160
p Value 0.00258711 0.66964815 0.00018319 0.0000000
AR (2) α2 0.19759
SEB 0.659442
T value 0.2996279
p Value 0.76538859
MA (2) θ2 0.17062 0.30202
SEB 0.142258 0.409353
T value 1.1993429 0.7377864
p Value 0.23455708 0.46322050
Constant Cte 124.42969 124.52640 125.53260 128.53887 129.22650
SEB 12.608189 12.601296 11.785537 6.5715235 4.8971088
T value 9.8689581 9.8820312 10.651411 19.559981 26.388326
p Value 0.00000000 0.00000000 0.00000000 0.00000000 0.00000000
AIC 688.86593 690.82312 688.77347 689.37103 693.59055 692.04831
SBC 697.9726 702.20645 695.60347 693.92437 698.14388 696.60164
Log likelihood 340.43297 340.41156 341.38674 342.68552 344.79527 344.02415
Error 28.034898 28.221721 28.214812 28.576249 29.443759 28.461048
SBC: Schwarz Bayesian criterion; AIC: Akaike criterion; ARIMA: autoregressive integrated moving average; SEB: Standard Error of B (B: Regression
Coefficient).

Table 3. ARIMA model parameters.

Estimate SE t Sig.
Sales (tons)-Model_1 No transformation Constant 125.524 12.466 10.069 0.000

AR Lag 1 0.908 0.095 9.537 0.000

MA Lag 1 0.639 0.162 3.938 0.000

ARIMA: autoregressive integrated moving average.

Figure 4. Sales, fit, LCL, and UCL. LCL: lower control limit; UCL: upper control limit.
Fattah et al. 7

Table 4. Forecast sales from January 2016 to October 2016.

Model 73 74 75 76 77 78 79 80 81 82
Sales-Model_1 Forecast 95.12 97.92 100.46 102.77 104.86 106.77 108.49 110.06 111.49 112.78
UCL 151.41 156.21 160.35 163.95 167.08 169.83 172.25 174.38 176.26 177.93
LCL 38.83 39.63 40.57 41.59 42.64 43.70 44.74 45.75 46.71 47.63
LCL: lower control limit; UCL: upper control limit.

Figure 5. Sales, fit, LCL, UCL, and forecasting. LCL: lower control limit; UCL: upper control limit.

Forecast Conclusion
After we have defined the most appropriate model of Demand forecasting is an important function of managing
demand in our case, we have to make the forecasting; to supply chain. Its integration with other business functions
do this and so to predict trends and develop forecast, we makes it one of the most important planning processes
used the IBM SPSS Forecasting. Table 4 and business can deploy for future. In this context, we devel-
Figure 5 present the results of the sales forecasts that we oped an ARIMA model to model the demand forecasting of
obtained by applying our model ARIMA (1, 0, 1) for the the finished product in a food manufacturing by using Box–
next 10 months from January 2016 to October 2016. Jenkins time series approach. The historical demand data
We can clearly see that the model chosen can be used for were used to develop several models and the adequate one
modeling and forecasting the future demand in this food was selected according to four performance criteria: SBC,
manufacturing, but each time we need to feed the historical AIC, standard error, and maximum likelihood. The model
data with the new data to enrich it in order to improve the that we selected and which minimizes the four previous
new model and forecasting. criteria is ARIMA (1, 0, 1). The results obtained proves
The forecasts obtained after modeling facilitated the that this model can be used for modeling and forecasting
decision on the production in this food company. In fact, the future demand in this food manufacturing; these results
the model enabled us to forecast the demand and make will provide to managers of this manufacturing reliable
accurate predictions. Once we obtain a demand forecast, guidelines in making decisions. As future work, we will
it will be much easier end very clear to make the right develop other models by using a combination of qualitative
production planning and thus eliminate big cost losses. and quantitative techniques to generate reliable forecasts
That will help us take right decisions related to supplying and increase the forecast accuracy. We will also try neural
raw materials and determination of daily production. More- network approach to compare it with ARIMA’s results in
over, that will affect the whole production process elimi- order to confirm the ANN’s strength in the food company.
nating then any kind of loss. Furthermore, we will make an ARIMA-radial basis
8 International Journal of Engineering Business Management

function (RBF) combination always to achieve the same 14. El Bahi Y, Ezzine L, Aman Z, et al. Modeling and forecasting
goal: high accuracy. of fuel selling price using time series approach: case study.
In: Proceedings of the international conference on control
Declaration of Conflicting Interests decision and information technologies, Thessaloniki, Greece,
The authors declared no potential conflicts of interest with respect 10–13 April 2018. IEEE.
to the research, authorship, and/or publication of this article. 15. Nogales F and Conejo A. Electricity price forecasting
through transfer function models. J Oper Res Soc 2006;
Funding 57(4): 1–7.
The authors received no financial support for the research, author- 16. Garcia R, Contreras J, Van Akkeren M, et al. A GARCH
ship, and/or publication of this article. forecasting model to predict day-ahead electricity prices.
IEEE Trans Power Syst 2005; 20(2): 867–874.
ORCID iD
17. Bozarth CB and Handfield RB. Introduction to operations
Zineb Aman http://orcid.org/0000-0001-6308-2507 and supply chain management, 4th ed. Raleigh: North Car-
olina State University, 2016.
References
18. Tsay RS. Time series and forecasting: brief history and future
1. Wisner JD, Tan KC, and Leong GK. Principles of supply research. J Am Stat Assoc 2000; 95: 638–643.
chain management: a balanced approach. Mason: Thomson 19. Ghobbar AA and Friend CH. Evaluation of forecasting
South-Western, 2011. methods for intermittent parts demand in the field of avia-
2. Dawei L. Fundamentals of supply chain management. Book- tion: a predictive model. Comput Oper Res 2003; 30:
boon.com, 2011. www.bookboon.com (accessed 15 March 2097–2114.
2017). ISBN: 978-87-7681-798-5. 20. Toktay B. Forecasting product returns. In: Guide Jr VDR, van
3. Baker M and Hart S. The marketing book, 7th ed. Amsterdam: Wassenhove LN (eds) Business aspects of closed-loop supply
Elsevier, 2016. Print ISBN: 9780415703765/10.4324/
chains. Pittsburgh: Carnegie Mellon University Press, 2003.
9781315890005.
21. Chopra S and Meindl P. Supply chain management. Strategy,
4. Gaafar L and Choueiki MH. A neural network model for
planning & operation. In: Das summa summarum des man-
solving the lot-sizing problem. Omega 2000; 28(2): 175–184.
agement. Gabler, 2007, pp. 265–275.
5. Sustrova T. An artificial neural network model for a whole-
22. Miller JJ, McCahon CS, and Miller JL. Foodservice forecast-
sale company’s order-cycle management. Int J Eng Bus
ing with simple time series models. J Hosp Tour Res 1991;
Manage 2016; 8: 2.
14: 9–21.
6. Prybutok VR, Yi J, and Mitchell D. Comparison of neural
23. Song H and Li G. Tourism demand modeling and forecast-
network models with ARIMA and regression models for pre-
ing—a review of recent research. Tour Manage 2008; 29:
diction of Houston’s daily maximum ozone concentrations.
203–220.
Eur J Oper Res 2000; 122(1): 31–40.
24. Willemain TR, Smart CN, and Schwarz HF. A new approach
7. Ho SL, Xie M, and Goh TN. A comparative study of neural
to forecasting intermittent demand for service parts invento-
network and Box-Jenkins ARIMA modeling in time series
ries. Int J Forecast 2004; 20: 375–387.
prediction. Comput Ind Eng 2002; 42(2–4): 371–375.
25. Taylor JW, de Menezes LM, and McSharry PE. A compari-
8. Kandananond K. A comparison of various forecasting meth-
ods for autocorrelated time series. Int J Eng Bus Manage son of univariate methods for forecasting electricity demand
2012; 4: 4. up to a day ahead. Int J Forecast 2006; 1: 1–16.
9. Aburto L and Weber R. Improving supply chain management 26. Weron R. Electricity price forecasting: a review of the state-
based on hybrid demand forecasts. Appl Soft Comput 2007; of-the-art with a look into the future. Int J Forecast 2014; 30:
7(1): 136–144. 1030–1081.
10. Mitrea CA, Lee CKM, and Wu Z. A comparison between 27. Mitsutaka M and Akira I. Examination of demand forecasting
neural networks and traditional forecasting methods: a case by time series analysis for auto parts remanufacturing.
study. Int J Eng Bus Manage 2009; 1: 11. J Remanuf 2015; 5: 1. DOI: 10.1186/s13243-015-0010-y.
11. Doulai P and Cahill W. Short-term price forecasting in elec- 28. Gardner ESJr. Exponential smoothing: the state of the art—
tric energy market. In: Proceedings of international power part II. Int J Forecast 2006; 22: 637–666.
engineering conference, (organisers, Nanyang Technical 29. Liu Q and Wang H. Research on the forecast and develop-
University, et al.), Grand Copthorne Waterfront, Singapore, ment of China’s public fiscal revenue based on ARIMA
17–19 May 2001, pp. 749–754. IEEE. model. Theor Econ Lett 2015; 5: 482–493.
12. Contreras J, Espinola R, Nogales F, et al. ARIMA models to 30. Shen S and Shen Y. ARIMA model in the application of
predict next-day electricity prices. IEEE Trans Power Syst Shanghai and Shenzhen stock index. Appl Math 2016; 7:
2003; 18(3): 1014–1020. 171–176.
13. Conejo A, Plazas M, Espinola R, et al. Day-ahead electricity 31. Kurawarwala AA and Matsuo H. Product growth models for
price forecasting using the wavelet transform and ARIMA medium-term forecasting of short life cycle products. Tech-
models. IEEE Trans Power Syst 2005; 20(2): 1035–1042. nol Forecast Soc Chang 1998; 57: 169–196.
Fattah et al. 9

32. Miller D and Williams D. Shrinkage estimators of time series 39. Bavaud F, Capel R, Crettaz de Roten F, et al. Guide de
seasonal factors and their effect on forecasting accuracy. Int J l’analyse statistique de donnees avec SPSS 6. Genève: Slat-
Forecast 2003; 19: 669–684. kine, 1996.
33. Hyndman RJ. The interaction between trend and seasonality. 40. Mélard G. Algorithm AS197: a fast algorithm for the exact
Int J Forecast 2004; 20(4): 561–563. likelihood of autoregressive-moving average models. J Royal
34. Box GEP and Jenkins G. Time series analysis, forecasting Stat Soc Ser C Appl Stat 1984; 33: 104–114.
and control. San Francisco: Holden-Day, 1970. 41. Mélard G and Roy R. Modèles de séries chronologiques avec
35. Yule GU. Why do we sometimes get nonsense-correlations seuils. Rev Stat Appl 1988; 4(36): 5–23.
between time series? A study in sampling and the nature of 42. Mélard G. Initiation à l’analyse des séries temporelles et à la
time series. J Royal Stat Soc 1926; 89: 1–64. prévision. Rev Mod 2006; 35: 82–129.
36. Wold H. A study in the analysis of stationary time series. 43. Box G and Jenkins G. Time series analysis, forecasting and
Stockholm: Almgrist & Wiksell, 1938. control, 3rd ed. San Francisco: Holden-Day, 1994.
37. Hanke JE and Reitsch AG. Business forecasting, 5th ed. Eng- 44. Brockwell PJ and Davis RA. Time series: theory and method.
lewood Cliffs. 1995. Berlin: Springer-Verlag, 1987.
38. Stafford J and Bodson P. L’Analyse multivariée avec SPSS. 45. Hamilton JD. Time series analysis. Princeton: Princeton
Quebec: Presses de l’Université de Québec [In French], 2006. University Press, 1994.