IOP Conference Series: Earth and Environmental Science

PAPER • OPEN ACCESS You may also like

- Establishing contact between cell-laden
Forecasting export demand for L-Lysine as animal hydrogels and metallic implants with a
biomimetic adhesive for cell therapy
feed product in PT X Indonesia supported implants
Julien Barthes, Angela Mutschler, Camille
Dollinger et al.
To cite this article: A Ihwah and R Syafira 2021 IOP Conf. Ser.: Earth Environ. Sci. 733 012047 - The application of material handling in fish
fillet production process of red snapper in
PT X
A L Rucitra

- Detection of DNA and poly-l-lysine using

View the article online for updates and enhancements. CVD graphene-channel FET biosensors
Aniket Kakatkar, T S Abhilash, R De Alba
et al.

This content was downloaded from IP address 210.187.177.97 on 14/03/2023 at 12:56

International Conference on Green Agro-industry and Bioeconomy IOP Publishing
IOP Conf. Series: Earth and Environmental Science 733 (2021) 012047 doi:10.1088/1755-1315/733/1/012047

Forecasting export demand for L-Lysine as animal feed

product in PT X Indonesia

A Ihwah and R Syafira

Department of Agro-industrial Technology, Faculty of Agricultural Technology, Universitas

Brawijaya, Malang, Indonesia
E-mail: [email protected]

Abstract. PT X Indonesia is one of the foreign companies from South Korea that engaged in
producing additional ingredients for animal feed. Located in Pasuruan - East Java, PT X is one
of the industries producing L-Lysine and L-Tryptophan essential amino acids. L-Lysine and L-
Tryptophan products are stimulants for animal feed which are exported to several countries. The
planning of the production process is based on requests from the headquarters of PT X Indonesia
in South Jakarta, Indonesia. To meet the needs of export products to several countries, PT X
requires careful planning in providing the products. One of the determining factors in planning
and providing products accurately, effectively, and efficiently is demand forecasting. This study
aimed to forecast the export demand of L-Lysine. Historical data used for forecasting was time-
series data from September 2019 to January 2020. The best ARIMA model combination chosen
for forecasting was (1,0,1) because both the p-values for the AR and MA models are significant
and it has the smallest Mean Square Error (MSE). Forecasting results use the ARIMA model
showed that L-Lysine export for four weeks was in a range of 2,596.47 tons to 2,597.5 tons with
a total of 10,387.9 tons and an average of 2,596.975 tons per week.

1. Introduction
PT. X Indonesia, located in Pasuruan - East Java, is one of the companies producing L-Lysine and L-
Tryptophan essential amino acids. L-Lysine and L-Tryptophan products are stimulants for animal feed
which are exported to several countries. To meet the needs of export products to several countries, PT
X requires careful planning in providing the products. One of the determining factors in planning and
providing products accurately, effectively, and efficiently is demand forecasting.
Demand forecasting can help PT. X Indonesia to predict market opportunities that are available in
the future. Demand forecasting is used by companies in predicting the number of product demand in the
next period. The purpose of a prediction is not to predict the conditions that occur in the future precisely,
but to look for information that will be used as a reference for production. Planning processes in
operations - e.g., capacity, production, inventory, and materials requirement plans - rely on a demand
forecast. The quality of these plans depends on the accuracy of this forecast [1].
The performance of candidate models is evaluated either on in-sample data, usually using appropriate
information, or by withholding a set of data points to create a validation sample (out-of-sample
evaluation, also known as cross-validated error) [2]. Demand forecasting has several stages that need to

Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
International Conference on Green Agro-industry and Bioeconomy IOP Publishing
IOP Conf. Series: Earth and Environmental Science 733 (2021) 012047 doi:10.1088/1755-1315/733/1/012047

be considered so that the forecasting process is effective and efficient. Stages of demand forecasting
begin with determining the objectives to be achieved from the implementation of forecasting, then
selecting the predicted object [3]. The next step is determining the time horizon of the data is retrieved.
The time horizon determines the forecasting method used to get the best results. The collected data is
then needed to be predicted. After the data is collected, the appropriate forecasting model is selected and
forecasting is finalized. Forecasting results are then analyzed and applied in the company.

2. Materials and Methods

Forecasting conducted in this study is quantitative forecasting using historical data regarding the number
of export demand for L-Lysine products every week. The forecasting method used was the ARIMA
method because ARIMA models are applied in some cases where data show evidence of non-
stationarity, where an initial differencing step (corresponding to the "integrated" part of the model) can
be applied one or more times to eliminate the non-stationarity [4]. In mathematics and statistics, a
stationary process (or a strictly stationary process or strongly stationary process) is a stochastic process
whose unconditional joint probability distribution does not change when shifted in time and
consequently, parameters such as mean and variance also do not change over time [5]. Therefore,
historical data will first be performed on the data stationarity test.
The historical data used was the export sales of L-Lysine 99% (L-Lysine HCL) products from the
first week of September 2019 to the third week of January 2020. Forecasting was conducted from the
fourth week of January 2020 to the second week of February 2020. The forecasting period is included
in the medium-term forecasting category because the period of data used was more than three months
[6].

3. Results and Discussion

Data on the number of L-Lysine 99% export sales (L-Lysine HCL) at PT X Indonesia, Pasuruan have a
horizontal data pattern because the data does not increase or decrease during a certain time. Horizontal
data patterns occur if the data fluctuates around a constant average value. A product whose sales have
not increased or decreased during a certain time is included in the type of horizontal pattern [7]. For
example, the sale of products that tend to be constant like groceries. Values of the average fluctuate
around 2,495 tons per week and do not indicate repetitive data movements from one period to the next.
The time-series graph from historical data on the number of export sales of L-Lysine 99% (L-Lysine
HCL) products can be seen in Figure 1. Furthermore, it will be seen whether each data has been
stationary towards the average and variance.

3500

3000
export of Lysine 99%

2500

2000

1500
2 4 6 8 10 12 14 16 18 20
week

Figure 1. A time series plot of export demand for L-Lysine.

3.1 Stationary test of variance

Stationary tests of variance were carried out using Box-Cox plots from the export of 99% L-Lysine (L-

2
International Conference on Green Agro-industry and Bioeconomy IOP Publishing
IOP Conf. Series: Earth and Environmental Science 733 (2021) 012047 doi:10.1088/1755-1315/733/1/012047

Lysine HCL). If the rounded value or lambda (λ) is equal to one, then the data is said to have been
stationary for variance [8]. Box-cox plots from 99% L-Lysine exports can be seen in Figure 2. The Box-
Cox plot showed a rounded value of one. It means that the data was in a stationary state of the variety
and does not need to be transformed.

Figure 2. Box-Cox plot export demand of L-Lysine.

Based on Figure 2, we can see the Lower Control Limit and Upper Control Limit of lambda, in which
these values were confidence intervals for the Box-Cox transformation. Those can be asymptotically
constructed using Wilks's theorem on the profile likelihood function to find all the possible values of λ
and showed in the graph by using vertical lines. The rounded value of lambda was more considered to
be used for data transformation than the estimated value because the rounded value was said to the
optimum value of lambda [9].

3.2 Stationary test of average

The stationary test of average was analyzed using the autocorrelation function graph. The
autocorrelation function (ACF) graph displayed the autocorrelation function and lag. At lag k, the bar
height represents the autocorrelation coefficients between series values that are k intervals apart [10]. If
the first three lags in the autocorrelation function plot have crossed the red line (significance limit), it
means that there is still autocorrelation and the data is not stationary in the average, so that it needs
differencing [11]. The plot of the L-Lysine 99% export autocorrelation function (ACF) can be seen in
Figure 3.

1,0

0,8

0,6

0,4
Autocorrelation

0,2

0,0

-0,2

-0,4

-0,6

-0,8

-1,0

2 4 6 8 10 12 14 16 18
Lag

Figure 3. ACF plot of export demand of L-Lysine.

3
International Conference on Green Agro-industry and Bioeconomy IOP Publishing
IOP Conf. Series: Earth and Environmental Science 733 (2021) 012047 doi:10.1088/1755-1315/733/1/012047

In Figure 3, no lag comes out of the red line, which is the red line is a confidence interval [12]. This
shows that the historical data on the number of L-Lysine export sales of 99% (L-Lysine HCL) has been
stationary to average. Moreover, the data was tested using the partial autocorrelation function (PACF)
plot as further testing to ensure that the data was stationary on average.
The PACF plot showed stationary patterns because no lag value exceeds the red line. These results
were similar to the previous autocorrelation function or ACF plot. This concludes that the data on the
amount of L-Lysine 99% export sales (L-Lysine HCL) was stationary to average. The results of the
partial autocorrelation function graph of the total L-Lysine export sales of 99% can be seen in Figure 4.

1,0

0,8

0,6
Partial Autocorrelation

0,4

0,2

0,0

-0,2

-0,4

-0,6

-0,8

-1,0

2 4 6 8 10 12 14 16 18
Lag

Figure 4. PACF plot.

3.3 Forecasting of export demand of L-Lysine

ARIMA model or Auto-Regressive Integrated Moving Average is a technique to find the most suitable
pattern of a group of data, by utilizing the past and present data of the dependent variable to make
accurate short-term forecasting [13]. The ARIMA model can predict irregular data due to data
fluctuations [14]. The determination of the ARIMA model is determined by three general components
namely (p, d, q), where p has the meaning of the autoregressive order, d means the level of differentiation
(differencing) and q is the order of the moving average [15]. The best model was said to have a p-value
less than 0.05 (p <0.05) which means that it was significant and it has the smallest mean square error
(MSE) value than the other models. The results of the significance test on a combination of the ARIMA
model can be seen in Table 1.
Table 1. Combination of ARIMA.
Model Type P-Value RMSE
(1,0,0) MA 1 0.000 803.807
(1,0,1) AR 1 0.000 649.819
MA 1 0.000
(0,0,1) AR 1 0.000 1,606.82

The best ARIMA model of the three models was ARIMA (1,0,1) with both AR and MA p-value of
zero so that it indicates the AR and MA models are significant, and it has the smallest RMSE value of
649.819. The square root of the sum of the square of the deviation of the predicted values from the
observed value dividing by their number of observation is known as the root mean square error [16].
Hence, demand forecasting using the ARIMA (1,0,1) model was done to find out the next period for the
next 4 weeks. The results of forecasting can be seen in Table 2.
Forecasting results using the ARIMA model with a range of 2,596.47 tons to 2,597.5 tons in one
week with a total of 10,387.9 tons and an average of 2,596.975 tons per week. To compare, the actual
data of L-Lysine exports 99% can be seen in Table 3. The actual data of L-Lysine exports obtained from

4
International Conference on Green Agro-industry and Bioeconomy IOP Publishing
IOP Conf. Series: Earth and Environmental Science 733 (2021) 012047 doi:10.1088/1755-1315/733/1/012047

companies tend to fluctuate with an average weekly sales of 2,563.75 tons and total sales of 10,255 tons
for 4 weeks. From these results, it can be seen the difference between the results of forecasting and
actual data was 33.225 tons on average. Forecasting results can be used by companies because it is not
much different when compared with the average value of the actual data. Therefore, the forecasting
model can be said to be sufficiently accurate to be used in forecasting L-Lysine export sales. However,
the results of forecasting tend to be constant whereas the actual data obtained tend to be volatile.
Forecasting results tend to be a constant possibility of a lack of data that represents for forecasting. The
more data used, the more accurate the data will be [17]. Actual data has fluctuations that cause irregular
data movements, caused by events that occur such as sudden export demand rises and competition so
that product demand also decreases. Animal feed sales tend to fluctuate depending on consumer demand
and the number of competitors of similar products [18].

Table 2. Forecasting of export demand of L-Lysine.

Period 2020 Week Forecasting Results (ton)
January 4 2,597.49
5 2,597.15
February 1 2,596.81
1 2,596.47
Total 10,387.9
Average 2,596.975

Table 3. Actual export demand of L-Lysine 99% from January 2020 to

February 2020.
Month Week Real export demand (ton)
January 4 1,847
5 2,076
February 1 2,997
2 3,308
Total 10,255
Average 2,563.75
Forecasting results can be used by companies because it is not much different when it was compared
to the average of actual data, so the forecasting model can be said to be accurate enough to be used in
forecasting L-Lysine export sales. Based on Singh et al [19], to find the best prediction model can be
done by looking at the error level generated, where if a smaller error value is the best model. Therefore,
may in another case, the reason why the forecasting result is significantly different from the actual
condition is because of the error value. Forecasting serves to make decisions in the company and
measure or estimate the situation in the future [20]. Forecasting results can be used as a benchmark to
calculate the amount of inventory in the coming period. So far, the company has only relied on the
number of requests from the central company. This can lead to an error in the form of a delay in the
arrival of raw materials caused by inaccurate time for the next period so that the production process can
be hampered.

4. Conclusions
The forecasting result of export demand of L-Lysine with a total of 10,387.9 tons and an average of
2,596.975 tons per week. The company can use these results to make a plan related to the needs of raw
materials, labor, machinery, and others. Product demand forecasting is very useful especially for
products that are made by order because the number of needs related to the production process can be
adjusted to the number of demands. Ordering of raw materials, the number of workers, setting working
hours, engine maintenance, and others can be well prepared if the company knows the predicted market

5
International Conference on Green Agro-industry and Bioeconomy IOP Publishing
IOP Conf. Series: Earth and Environmental Science 733 (2021) 012047 doi:10.1088/1755-1315/733/1/012047

demand for its products. If properly prepared, constraints such as lack of raw materials, the amount of
production that exceeds production capacity, or damage to the machine will be controlled. Thus,
consumers will feel satisfied not only with product quality but also with the services provided by the
company.

References
[1] Fotios P, Nikolaos K, Konstantinos N, Enno S 2018 Judgmental selection of forecasting models
J. Operation Manag 60 34 – 46
[2] Ord J K, Fildes R, Kourentzes N 2017 Principles of business forecasting (Indiana: Wessex Press
Publishing Co)
[3] Eunike A, Nasir W S, Rahmi Y, Ihwan, H, Ro P L, Angga A F 2018 Perencanaan produksi dan
pengendalian persediaan (Production planning and inventory control) (Malang: UB Press) [In
Indonesian]
[4] Asteriou D H, Stephen G 2011 ARIMA Models and the box–jenkins methodology Applied
Econometrics (Second ed.) (New York: Palgrave MacMillan) pp 265–286
[5] Gagniuc O A 2017 Markov chains: from theory to implementation and experimentation (USA:
John Wiley & Sons) pp 1–256
[6] Pianda D 2018 Optimasi perencanaan produksi pada kombinasi produk dengan metode linear
(Production planning optimization on product combination using linier method) (Sukabumi:
CV Jejak) [In Indonesian]
[7] Christian E O 2012 Modified trend and seasonal time series analysis for operations: a case study
of soft drink production Int'l J. of Eng Rsch in Africa 7 63 – 72
[8] Montgomery DC, Jennings C L, Kulahci L 2008 Introduction to time series analysis and
forecasting Evaluating and monitoring forecasting model performance (New York: John
Wiley & Sons) pp 49 – 60
[9] Abramovich, Felix,Ya'akov Ritov 2013 Statistical theory: a concise introduction (Florida: CRC
Press)
[10] Purnomo R A 2017 Analisis Statistik Ekonomi Dan Bisnis dengan SPSS (Analysis of statistics
economics and business using SPSS) (Ponorogo: Wade Group) [In Indonesian]
[11] Pamungkas M, Arief W 2018 Aplikasi metode ARIMA Box-Jenkins untuk meramalkan kasus
DBD di Provinsi Jawa Timur (The Application of ARIMA Box-Jenkins method to forecast
DBD cases in East Java province) The Indonesian J. of Public Health 13 2 181-194 [In
Indonesian]
[12] Moore D 2016 Applied survival analysis using R (New York: Springer)
[13] Swain S, Nandi S, Patel P 2018 Development of an ARIMA Model for Monthly Rainfall
Forecasting over Khordha District, Odisha, India Recent Findings in Intelligent Computing
Techniques (New York: Springer) chapter 708 pp 325 – 331
[14] Castillo O, Melin P 2018 Fuzzy logic augmentation of neutral and optimization algorithms:
theoretical aspects and real applications (New York: Springer)
[15] Hyndman RJ, Athanasopoulos G 2018 Forecasting: principles and practice second edition
(Melbourne: Otexts)
[16] Jiban C P, Shahidul H, Mohammad M R 2013 Selection of best ARIMA model for forecasting
average daily share price index of pharmaceutical company in Bangladesh: a case study on
square pharmaceutical company Ltd J. of Man and Business Rsch Fin 13 3 1.0
[17] Maricar MA 2019 Analisa perbandingan nilai akurasi moving average dan exponential smoothing
untuk sistem peramalan pendapatan pada perusahaan XYZ (Comparison analysis of accuration
value using moving average and exponential smoothing to forecasting system of income in
XYZ company) J. Sistem dan Informatika 1 3 36-46
[18] Hadinata 2017 Penerapan metode exponential smoothing dalam peramalan biaya pengolahan
peternakan ayam (Implication of exponential smoothing method in forecasting of production
cost in chicken farm) J. Siskom 1 2 1-5 [In Indonesian]

6
International Conference on Green Agro-industry and Bioeconomy IOP Publishing
IOP Conf. Series: Earth and Environmental Science 733 (2021) 012047 doi:10.1088/1755-1315/733/1/012047

[19] Dharm S, Naveen C, Jully S 2013 Analysis of data mining classification with decision tree
technique Global J. of Comp Sci and Tech 13 13 1 – 7
[20] Herjanto E 2015 Manajement Operasi Edisi Tiga (Operation Management) ( J a k a r t a :
P enerbit Grasindo)