Page 1 of 9
Advances in Social Sciences Research Journal – Vol. 11, No. 11
Publication Date: November 25, 2024
DOI:10.14738/assrj.1111.17941.
Shafie, N. A., Borhan, N., Ariffin, N. A. N., Isa, K. A. M., & Ghani, N. A. M. (2024). Modeling and Forecasting of Total Supply Malaysia’s
Centrifugal Sugar Using ARIMA Model. Advances in Social Sciences Research Journal, 11(11). 238-246.
Services for Science and Education – United Kingdom
Modeling and Forecasting of Total Supply Malaysia’s Centrifugal
Sugar Using ARIMA Model
Nur Amalina Shafie
Mathematical Sciences Studies, College of Computing, Informatics and
Mathematics, Universiti Teknologi MARA Negeri Sembilan Branch,
Seremban Campus, 70300 Seremban, Negeri Sembilan, Malaysia
Nurbaizura Borhan
School of Mathematical Sciences, College of Computing,
Informatics and Mathematics, Digital Innovation
Noor Amalina Nisa Ariffin
Mathematical Sciences Studies, College of Computing, Informatics and
Mathematics, Universiti Teknologi MARA Negeri Sembilan Branch,
Seremban Campus, 70300 Seremban, Negeri Sembilan, Malaysia
Khairil Anuar Md Isa
Department of Basic Sciences, Faculty of Health Sciences,
Universiti Teknologi MARA Selangor Branch,
Puncak Alam Campus, 42300 Selangor, Malaysia
Nor Azura Md Ghani
Mathematical Sciences Studies, College of Computing, Informatics and
Mathematics, Universiti Teknologi MARA Negeri Sembilan Branch,
Seremban Campus, 70300 Seremban, Negeri Sembilan, Malaysia
ABSTRACT
Centrifugal sugar is essential to Malaysia's food sector and culinary scene because
it is high-quality, consistent, and convenient for a variety of food and beverage
applications. Customers, chefs, and food manufacturers who are looking for
dependable sweetening solutions for their regular cooking and dining experiences
choose it because of its broad availability and adaptability. To support market
stability, price control, production scheduling, trade policy development, risk
management, policy formation, and consumer welfare, it is essential to forecast the
centrifugal sugar supply in Malaysia. Autoregressive Integrated Moving Average
(ARIMA) model is one of Box-Jenkins model which is common time series
forecasting methodology that is widely utilised in various sectors including
economics, finance, and business. Therefore, this research interest to model and
forecast the total supply of centrifugal sugar in Malaysia using ARIMA model. This
research used data from the IndexMundi website and analysed by using R software.
This research found that the best model to forecast the total supply Malaysia’s
centrifugal sugar is ARIMA (1,1,2). The total supply of Malaysia’s centrifugal sugar
will increase in 6 years ahead.
Page 2 of 9
239
Shafie, N. A., Borhan, N., Ariffin, N. A. N., Isa, K. A. M., & Ghani, N. A. M. (2024). Modeling and Forecasting of Total Supply Malaysia’s Centrifugal
Sugar Using ARIMA Model. Advances in Social Sciences Research Journal, 11(11). 238-246.
URL: http://dx.doi.org/10.14738/assrj.1111.17941
Keywords: Centrifugal sugar, ARIMA, total supply, forecasting.
INTRODUCTION
Centrifugal sugar is included in the production and trading of sugar which is increased
significantly worldwide; the top exporting countries are Brazil, Thailand, and India [1-2].
However, Malaysia and other ASEAN countries face challenges in the sugar industries despite
their significant role in global trade [3]. Agidi and Igbeka [4] offers a particular illustration of a
small-scale sugar centrifuge that would be able to supplement Malaysia's overall centrifugal
sugar supply. Centrifugal sugar refers to sugar that has been extracted from a liquid by use of a
centrifugal machine. It is the result of the machine’s successful separation of sugar crystals
through the application of centrifugal force. Using centrifugal force, sugar crystals are separated
from the surrounding liquid, such as molasses or syrup. In the sugar processing sector,
centrifugal sugar is essential, especially for small-scale operations. Although it can cause some
crystal disintegration, it helps eliminate contaminants in sugar [5]. It has been discovered that
applying centrifugal force to the solute extraction from sugar beetroot tissue greatly improves
the extraction process [6].
Hoh [7] and Yuttitham et al. [8] mentioned that Malaysia's sugar production is primarily
centrifugal sugar, with the country being only about 5 percent self-sufficient in domestic sugar
production. This indicates that most of Malaysia's sugar production consists of centrifugal
sugar, with a small percentage being domestically produced. Malaysia's cane production is
predicted to reach 800 TMT in 2006 due to improved growth conditions and dry weather [9].
In 2006 and 2007, a 3% rise in domestic sugar consumption is anticipated. Tarimo and
Takamura [10] mentioned that in terms of local sugar production, Malaysia is just
approximately 5% self-sufficient.
Numerous studies have effectively employed the Auto-Regressive Integrated Moving Average
(ARIMA) model to forecast sugar production. ARIMA models are widely used in time series
analysis to understand and forecast data patterns. Integrated (I), Auto-Regressive (AR), and
Moving Average (MA) are the three main components of the ARIMA model. In order to
anticipate Malaysia's total supply of centrifugal sugar using ARIMA, the analysis of historical
supply data will determine the essential ARIMA parameters (p, d, q) based on the
autocorrelation and partial autocorrelation functions. These parameters reflect the number of
moving average terms, differencing, and autoregressive terms, respectively. Once fitted to the
data, the ARIMA model can be used to predict future levels of supply. Numerous studies have
shown that ARIMA models perform well when used for time series analysis. It was discovered
by Zhang [11] that hybrid ARIMA and neural network model increased predicting accuracy.
Fattah et al. [12] forecasted demand in a food industry using ARIMA models with success; the
chosen model was verified against historical data. Both Ariyo et al. [13] and Mondal et al. [14]
discovered that ARIMA models could accurately forecast stock values; Mondal et al. [14] also
pointed out that these models might be used for short-term forecasting. All these studies
demonstrate how adaptable and trustworthy ARIMA models are for a range of forecasting
applications.
The ARIMA model would specifically be used for the sugar supply in Malaysia to examine past
supply data, spot trends, seasonality, and other supply-influencing factors, and then project
future supply levels based on these patterns. Putra et al. [15] created a forecasting system for
Page 3 of 9
240
Advances in Social Sciences Research Journal (ASSRJ) Vol. 11, Issue 11, November-2024
Services for Science and Education – United Kingdom
Indonesian sugar production, Wardah et al. [16] combined demand forecasting and inventory
control for coconut sugar. The ARIMA model was employed by Devaki and Mohideen [17] and
Muhammad et al. [18] to predict the production of sugarcane in Tamil Nadu and Pakistan,
respectively. By using the ARIMA model to anticipate maize production in India, Sharma et al.
[19] demonstrated how adaptable this forecasting method is for a variety of crops and nations.
Tripathi et al. [20] used the ARIMA model as well to estimate Odisha's rice production and
compared the results with those for the entire country of India. Kumar et al. [21] used Box- Jenkin’s ARIMA model to forecast sugarcane production in India. Saleem et al. [22] also used
the ARIMA model to forecast the production of centrifugal sugar in Pakistan. This research
focuses on modelling and forecasting the total supply of Malaysia’s centrifugal sugar by using
the ARIMA model.
METHODOLOGY
This research used annual data total supply of centrifugal sugar in Malaysia from 1972 to 2023
from the IndexMundi website. The estimating and evaluation portions of the data were
separated into two groups. In the estimating stage, centrifugal sugar measurements from 1972
to 2002 were used as training data. Meanwhile, the evaluation portion employed centrifugal
sugar data from 2003 to 2023. This research used descriptive analysis and Autoregressive
Integrated Moving Average (ARIMA) model. Descriptive statistics are used to get the trend of
the raw data. While the ARIMA model was focused on time series prediction.
ARIMA Model
The ARIMA model is one type of Box-Jenkins model that is used to predict future time series
data based on historical time series data. The ARIMA model is a generalization of the ARMA
model. The ARMA model is used for stationary data. A stationary zero mean ARMA (p, q) model
for a time series {Xt} is defined as Equation 1.
Xt = φ1Xt−1 + φ2Xt−2 + ⋯ + φpXt−p + εt + θ1εt−1 + θ2εt−2 + ⋯ + θqεt−q (1)
where
• Xt is the value of time series at time t.
• ∅1, ∅2, ⋯, ∅p are the parameters of the autoregressive part of the model.
• θ1, θ2, ⋯, θq are the parameters of the moving average part of the model. {εt} is the white
noise error term at time t, with zero mean and constant variance.
When a time series is not stationary, then differencing operations are applied with the
appropriate lag to achieve stationarity then is called the ARIMA model. The ARMA model has
two order parameters which are p and q. While, the ARIMA model has three order parameters
which are p, d, and q. p refers to the order of the autoregressive process, d refers to the order
of integration or differencing and q refers to the order of the moving average method used in
the model.
The raw data is separated into two parts which are training data and testing data. The training
data is used to fit the model while the testing data is used to evaluate the model’s performance.
The training data is identified whether is stationary or not by using a unit root test. The data is
considered stationary when its mean, variance, and autocovariance are all-time invariant. This
Page 4 of 9
241
Shafie, N. A., Borhan, N., Ariffin, N. A. N., Isa, K. A. M., & Ghani, N. A. M. (2024). Modeling and Forecasting of Total Supply Malaysia’s Centrifugal
Sugar Using ARIMA Model. Advances in Social Sciences Research Journal, 11(11). 238-246.
URL: http://dx.doi.org/10.14738/assrj.1111.17941
research used the Augmented Dickey-Fuller (ADF) test as Equation 2 to determine whether the
data is stationary or not by following the hypothesis below:
• H0 ∶ The series has a unit root
• H1 ∶ The series is stationary
∆yt = α + βt + γyt−1 + δ1∆yt−1 + δ2∆yt−2 + ⋯ + δp∆yt−p + εt (2)
where
• ∆ is the first difference operator.
• yt is the time series.
• t is the time trend.
• εt is the error term.
The data is stationary if the p-value is less than α=0.05, the null hypothesis is rejected. This
means that the unit root is absent from the series. If the series is not stationary, it should be
adjusted to become so, and differentiation is a typical way to do this. A data is differencing by
subtracting the current and previous values d times. When the stationarity assumption is
violated, differencing is frequently utilized to stabilize the series. Autocorrelation Function
(ACF) correlogram also can help to identify stationary data. The data is stationary if the data
fluctuates randomly around some fixed values.
The ACF correlogram is used to determine the q order by counting the number of significant
lags. The Partial Autocorrelation Function (PACF) correlogram is used to determine the p order
by counting the number of significant lags. After that, the residuals of suggestion models are
checked to either fulfill the assumption of white noise or not by using the Ljung Box test as
Equation 3 by following the hypothesis below:
• H0 ∶ The residuals have white noise (there is no autocorrelation)
• H1 ∶ The residuals have no white noise (there is significant autocorrelation)
Q = n(n + 2) ∑
r̂k
2
n−k
m
k=1
(3)
where
• n is the number of observations
• m is the number of lags being tested
• rk̂is the sample autocorrelation at lag k.
If the result has a p-value more than α = 0.05, so the residuals have white noise since it failed to
reject the null hypothesis. The models that have white noise are chosen. After that, the
performance of forecasting models can be evaluated using forecasting accuracy measures such
as Mean Square Error (MSE) and Root Mean Square Error (RMSE). The formula for the
measures is shown in Equations 4 and 5. The best forecasting model is found at the smallest
MSE and RMSE.
MSE = ∑
(σt
2−σ̂t
2
)
2
N
N
t=1
(4)
Page 5 of 9
242
Advances in Social Sciences Research Journal (ASSRJ) Vol. 11, Issue 11, November-2024
Services for Science and Education – United Kingdom
RMSE = √∑
(σt
2−σ̂t
2
)
2
N
N
t=1
(5)
where
• N is the number of total observations t is the t-th observation in out-of-sample
• σt
2 and σ̂t
2 correspond to the actual and predicted condition variance at time t,
respectively.
RESULTS AND DISCUSSION
Raw Data
Based on Figure 1, the trend of centrifugal sugar was on the rise. It means the total supply of
centrifugal sugar in Malaysia from 1972 to 2023 is increasing.
Figure 1: Trend analysis for total supply Malaysia’s centrifugal sugar
Test of Stationary
As mentioned above the data were separated into two groups which are from 1972 to 2002
were used as training data and data from 2003 to 2023 as testing data. For the test of stationary,
training data were used. The ACF of the data, as seen in Figure 2, displayed a declining pattern,
supporting the idea that the data was not stationary. Additionally, the ADF test in Table 1 also
verifies that the data is not stationarity since the p-values above α = 0.05 (0.4559). As a result,
there is insufficient proof that the data has a unit root at the 5% level of significance. Hence, the
data is not stationary.
Figure 2: ACF correlogram of original data
Table 1: Unit root test of original data
Test p-value
Augmented Dickey-Fuller (ADF) 0.4559
Page 6 of 9
243
Shafie, N. A., Borhan, N., Ariffin, N. A. N., Isa, K. A. M., & Ghani, N. A. M. (2024). Modeling and Forecasting of Total Supply Malaysia’s Centrifugal
Sugar Using ARIMA Model. Advances in Social Sciences Research Journal, 11(11). 238-246.
URL: http://dx.doi.org/10.14738/assrj.1111.17941
The differencing technique was used to fulfill the stationary analysis assumptions. ACF of first- order differencing is displayed in Figure 3. The initial differencing between data is stationary
since the plot does not exhibit a declining pattern. The null hypothesis was rejected when the
results of the ADF test in Table 2 showed p-values of 0.04526, which are less than α = 0.05.
Consequently, there is sufficient evidence to conclude that the data lacks a unit root at the 5%
significance level. The initial set of differencing data is therefore stationary.
Figure 3: ACF correlogram of first
differencing
Figure 4: PACF correlogram of first
differencing
Table 2: Unit root test after diffferencing
Test p-value
Augmented Dickey-Fuller (ADF) 0.04526
Modelling
Figures 3 and 4's ACF and PACF can be used to create an ARIMA model. The ACF indicates
testing MA(1) since it displays one notable spike at lag 0. In the meantime, PACF displays a
single notable peak at lag 8. It therefore recommends testing the AR (1) model. For this
investigation, a total of six reasonable ARIMA models were examined.
Table 3 shows that all suggested models have p-values more than α = 0.05 except ARIMA (2,1,1).
It means all the suggested models except ARIMA (2,1,1) are white noise since all the suggested
models failed to reject the null hypothesis.
Table 3: Portmanteau test
Model Ljung-Box Test (p-value)
ARIMA(0,1,1) 0.3214
ARIMA(1,1,0) 0.3875
ARIMA(1,1,1) 0.149
ARIMA(1,1,2) 0.3314
ARIMA(2,1,1) 0.03525
ARIMA(2,1,2) 0.0758
Forecast
The values for the forecasting accuracy metrics are displayed in Table 4. The lowest values of
MSE and RMSE values is ARIMA (1,1,2) when compared to other models. As a result, this model
might be regarded as the series' best forecasting model.
Page 7 of 9
244
Advances in Social Sciences Research Journal (ASSRJ) Vol. 11, Issue 11, November-2024
Services for Science and Education – United Kingdom
Table 4: Forecast accuracy
Model MSE RMSE
ARIMA(0,1,1) 1931872.7631 1389.9183
ARIMA(1,1,0) 2024512.3923 1422.8536
ARIMA(1,1,1) 1931820.0783 1389.8993
ARIMA(1,1,2) 1923849.0835 1387.0288
ARIMA(2,1,2) 1925908.6152 1387.7711
After selecting the ARIMA (1,1,2) model, the model was used to forecast 6 years ahead. Table 5
shows the forecast value and Figure 5 shows the graph of forecast value.
Table 5: Forecast value
Year Forecast value (1000 MT)
2024 2725.946
2025 2874.475
2026 2857.472
2027 2859.418
2028 2859.195
2029 2859.221
Figure 5: Forecasting plot
CONCLUSIONS
Centrifugal sugar is essential to Malaysia's food sector. Centrifugal sugar is sugar that has been
extracted from liquid using a centrifugal machine. While non-centrifugal cane sugar (NCS) is a
technical term for conventional raw sugar made by evaporating water from sugarcane juice.
The sugar industry in Malaysia is facing several challenges, including those related to market
dynamics and technology utilization. Encompassing economic, social, and environmental
aspects, centrifugal sugar forecasting is very crucial to be taken into account. The accurate
forecasting techniques can help to ensure that there will be a sufficient supply to meet the
demand of the population. In addition, forecasting supply accurately will help in stabilizing food
prices by preventing overproduction or underproduction that somehow will lead to price
volatility. Therefore, this research has been devoted to predicting the total supply of centrifugal
sugar prediction in Malaysia using the ARIMA model approach. Using the ARIMA model
approach will involve several steps. Starting from checking stationary to model diagnostics and
then forecasting process. It is indeed a powerful forecasting tool and can be effectively applied
for centrifugal sugar supply data in predicting future trends which helps in efficient future
planning and decision making. This prediction of centrifugal sugar plays a critical role in
enhancing economic stability, public health, environmental sustainability, and overall society
Page 8 of 9
245
Shafie, N. A., Borhan, N., Ariffin, N. A. N., Isa, K. A. M., & Ghani, N. A. M. (2024). Modeling and Forecasting of Total Supply Malaysia’s Centrifugal
Sugar Using ARIMA Model. Advances in Social Sciences Research Journal, 11(11). 238-246.
URL: http://dx.doi.org/10.14738/assrj.1111.17941
well-being. By leveraging accurate forecasts, it can help stakeholders to make informed
decisions that not only benefit the food sector but also the broader economy and society. This
research found that the best model to forecast centrifugal sugar is ARIMA (1,1,2) since it has
the lowest accuracy error. This research also found that the values of six years ahead have
gradually increased pattern.
ACKNOWLEDGEMENT
We would like to thank Universiti Teknologi MARA for their financial support through (600-
RMC/GPM LPHD 5/3 (136/2021)).
References
[1] Mariotti, M., & Lucisano, M. (2014). Sugar and sweeteners. Bakery products science and technology, 199-
221.
[2] Godshall, M. A. (2007). Sugar and other sweeteners. In Kent and Riegel’s Handbook of Industrial Chemistry
and Biotechnology (pp. 1657-1693). Boston, MA: Springer US.
[3] Solomon, S., Swapna, M., Xuan, V. T., & Mon, Y. Y. (2016). Development of sugar industry in ASEAN
countries. Sugar Tech, 18, 559-575.
[4] Agidi, G., & Igbeka, J. C. (2003). Development and performance of a sugar centrifuge. Sugar tech, 5, 131-
136.
[5] Jullienne, L. M. S. A. (1983). Washing sugar in batch A-centrifugals. Proceedings of The South African Sugar
Technologists' Association-June, 43.
[6] EL‐BELGHITI, K. A. M. A. L., Rabhi, Z., & Vorobiev, E. (2005). Effect of centrifugal force on the aqueous
extraction of solute from sugar beet tissue pretreated by a pulsed electric field. Journal of Food Process
Engineering, 28(4), 346-358.
[7] Hoh, R. (2006). Malaysia Sugar Annual 2006. USDA Foreign Agricultural Services Global Agriculture
Information Network (GAIN) Report.
[8] Yuttitham, M., Gheewala, S. H., & Chidthaisong, A. (2011). Carbon footprint of sugar produced from
sugarcane in eastern Thailand. Journal of Cleaner Production, 19(17-18), 2119-2127.
[9] Li, Y. R., & Yang, L. T. (2015). Sugarcane agriculture and sugar industry in China. Sugar Tech, 17(1), 1-8.
[10] Tarimo, A. J., & Takamura, Y. T. (1998). Sugarcane production, processing and marketing in Tanzania.
African Study Monographs, 19(1), 1-12.
[11] Zhang, G. P. (2003). Time series forecasting using a hybrid ARIMA and neural network model.
Neurocomputing, 50, 159-175.
[12] Fattah, J., Ezzine, L., Aman, Z., El Moussami, H., & Lachhab, A. (2018). Forecasting of demand using ARIMA
model. International Journal of Engineering Business Management, 10, 1847979018808673.
[13] Ariyo, A. A., Adewumi, A. O., & Ayo, C. K. (2014, March). Stock price prediction using the ARIMA model. In
2014 UKSim-AMSS 16th international conference on computer modelling and simulation (pp. 106-112).
IEEE.
[14] Mondal, P., Shit, L., & Goswami, S. (2014). Study of effectiveness of time series modeling (ARIMA) in
forecasting stock prices. International Journal of Computer Science, Engineering and Applications, 4(2), 13.
Page 9 of 9
246
Advances in Social Sciences Research Journal (ASSRJ) Vol. 11, Issue 11, November-2024
Services for Science and Education – United Kingdom
[15] Putra, J. A., Basbeth, F., & Bukhori, S. (2019, September). Sugar production forecasting system in PTPN XI
semboro jember using autoregressive integrated moving average (ARIMA) method. In 2019 6th
International Conference on Electrical Engineering, Computer Science and Informatics (EECSI) (pp. 448-
453). IEEE.
[16] Wardah, S., Nurhasanah, N., & Sudarwati, W. (2023). Integration models of demand forecasting and
inventory control for coconut sugar using the ARIMA and EOQ modification methods. Jurnal Sistem dan
Manajemen Industri, 7(2), 127-138.
[17] Devaki, C., & Mohideen, A. K. (2022). A sugarcane production in tamilnadu using a comparison of arima,
state space and linear mixed models. International Journal of Mechanical Engineering, 970-974.
[18] Muhammad, F., Javed, M. S., & Bashir, M. (1992). Forecasting sugarcane production in Pakistan using
ARIMA Models. Pak. J. Agric. Sci, 9(1), 31-36.
[19] Sharma, Pawan Kumar, Sudhakar Dwivedi, Lyaqat Ali, and R. K. Arora. "Forecasting maize production in
India using ARIMA model." Agro-Economist 5, no. 1 (2018): 1-6.
[20] Tripathi, Rahul, A. K. Nayak, R. Raja, Mohammad Shahid, Anjani Kumar, Sangita Mohanty, B. B. Panda, B.
Lal, and Priyanka Gautam. "Forecasting rice productivity and production of Odisha, India, using
autoregressive integrated moving average models." Advances in Agriculture 2014 (2014).
[21] Kumar, Manoj, and Madhu Anand. "An application of time series ARIMA forecasting model for predicting
sugarcane production in India." Studies in Business and Economics 9, no. 1 (2014): 81-94.
[22] Saleem, Muhammad Kashif, Parwarasha Nazir, and Fahim Nazeer. "Forecasting the Production of
Centrifugal Sugar in Pakistan using ARIMA Modelling." Journal of Statistics 27 (2023): 96.
[23] IndexMundi (2024).
https://www.indexmundi.com/agriculture/?country=my&commodity=centrifugalsugar&graph=total- supply