Scientific Journal Warsaw University of Life Sciences – SGGW

Problems of World Agriculture volume 22 (XXXVII), number 2, 2022: 17-29

DOI: 10.22630/PRS.2022.22.2.6
Mary Cris F. Pleños1
Visayas State University, Philippines

Time Series Forecasting Using Holt-Winters Exponential

Smoothing: Application to Abaca Fiber Data
Abstract. This study utilized the data on abaca fiber production and used Holt-Winters model to forecast
the abaca fiber production since the studied variable is characterized by a fairly strong intensity of
seasonality. For the construction of forecasts, additive and multiplicative models were used. The most
accurate forecasts were selected on the basis of Mean Square Error, Root Mean Square Error, Mean
Absolute Percentage Error, and Mean Absolute Scaled Error. It was found that the multiplicative
method had a higher accuracy, hence it was utilized to forecast the production for the next three years.
According to the findings, the anticipated fiber production for 2021-2023 showed an increase up to the
second quarter, but then declining afterwards.

Key words: abaca fiber, time series, forecasting, Holt-Winters method

JEL Classification: Q1, C1

Introduction

Abaca (Musa textilis) is a close relative of the banana. Abaca is a leaf fiber, composed
of long slim cells that form part of the leaf's supporting structure. This plant is native to the
Philippines and widely distributed in the humid tropics. It is also cultivated in other Southeast
Asian countries with Ecuador as the second largest producing country next to the Philippines
(Food and Agriculture Organization, 2021).
The Philippine abaca industry was a key player in the global abaca market (Research
and Markets, 2015). The Philippines supplies 80 percent of the world's abaca need. It has
become a source of employment for more than 1.5 million Filipinos who rely on it for a
living, either directly or indirectly (Far Eastern Agriculture, 2020).
According to Food and Agriculture Organization (2021), abaca is prized for its great
mechanical strength, resistance to saltwater damage, and long fiber length. It is used for
specialty papers such as currency notes, tea and coffee bags, vacuum bags, cigarette filter
paper, sausage casing paper, and high-quality writing paper. It is also used to make twines,
ropes, and fishing lines and nets. Abaca has a high potential to substitute glass fibers in
multiple automotive parts.
Almost all abaca produced is exported, primarily to Europe, Japan, and the United
States. With this, there is a growing demand for abaca on the global market yet the current
production level is not sufficient to meet this. Weather disturbances and the threat of bunchy
top disease, which has wiped out the abaca sector, are among the reasons for the low
production. In fact, there is a demand-supply gap in 2019 estimated to be 25,000 metric tons
(Waller & Wilsby, 2019). With the rising demand for abaca fiber and the existing challenges

1
National Abaca Research Center, Visayas State University, e-mail: mc.plenos@vsu.edu.ph;
https://orcid.org/0000-0002-8378-2663

https://prs.sggw.edu.pl/index/
18 M.C.F. Pleños

being faced by the abaca sector, proper planning by responsible agencies is required to
improve its production and be able to meet the global demand for abaca. Appropriate
planning is required to organize their work and provide the resources they will need in the
future. With this, production estimates should be known through statistical forecasting
techniques in order to plan appropriately. We can provide reliable inputs to planning when
we have accurate predictions or trends on abaca production.
There are a number of time series forecasting methods such as ARIMA, TBAT,
Facebook’s Model Prophet and many others. The choice of algorithm depends on the problem
since not all procedure are applicable to the available data set. The appropriate time series
model will depend upon the characteristics of data.
This paper develops a predictive model of Holt-Winters method after testing the
accuracy of additive and multiplicative HW models. Using the selected model, the quarterly
production of abaca for the years 2021 to 2023 is predicted. The goal of the study is to provide
information on future abaca production trends for the next three years using Holt-Winters
that the government bodies may utilize as they decide on policy decisions that will help the
abaca sector improve. As a result, this research helps to establish more initiatives that may
also have an impact on the welfare of abaca farmers. This study also contributes to the
growing literature of abaca studies.

Literature Review

Time Series Analysis is the most widely used field of data science and machine learning,
it is a type of predictive analysis that forecasts the value of a variable in future occurrences
based on history (Analytics Vidhya, 2022). A time series is a collection of observations of
well-defined data items obtained through repeated measurements over time (Australian
Bureau of Statistics, 2022). There are a number of time series forecasting methods but the
choice of algorithm depends on the business problems and the data set at hand. Not all
procedure are applicable to the available data set. The appropriate time series model will also
depends upon the characteristics of data.
ARIMA and SARIMA models was used to predict influenza in 2012 (Song, 2016).
Gecili et al. (2021) applied four time series models (Holt, ARIMA, TBATS, and cubic
smoothing spline model) to publicly available daily COVID-19 data for both the USA and
Italy. Liu et al. (2020) used Bayesian time-series framework to predict the number of
COVID-19 infection cases in USA. Satrio et al. (2021) utilized Facebook’s Prophet Model
and the ARIMA model to forecast the trend of the COVID19 diseases in Indonesia. The
COVID-19 pandemic in Saudi Arabia was analyzed using modified singular spectrum
analysis or SSA (Alharbi, 2021). Least Square Method, Moving Average Method, Single
Exponential Method, Double Exponential Method and Winter’s Method were used in
forecasting agricultural products prices (Ruekkasaem L. and Sasananan M., 2018)
However this paper mainly focuses on using Holt-Winters Exponential Smoothing in
time series forecasting. Holt-Winter’s Exponential Smoothing as named after its two
contributors: Charles Holt and Peter Winter’s is one of the oldest time series analysis
techniques which takes into account the trend and seasonality while doing the forecasting
(Solar Winds Worldwide, 2021). HW method is used for short term forecasts (Goodwin,
2010). Depending on the type of seasonality, it can either be additive or multiplicative.
According to Mgale et al. (2021), forecasts will depend on three components of seasonal time
Time Series Forecasting Using Holt-Winters Exponential Smoothing: Application to Abaca … 19

series: its level, its trend and its seasonal coefficient. Exponential smoothing is a method to
smooth a time series where it allocates exponentially decreasing weights and values in
opposition to historical data to lessen the value of the weights for the bygone data.
Exponential smoothing can be classified into three types. While the simple or single
exponential smoothing time series forecasting for uni-variate data does not have any
systematic structure with no trend and seasonality. In this case, only single parameter α is
used as a smoothing factor that lies between 0 and 1. A smaller value α designate slow
learning, takes more past observations for forecasting and a larger value designate faster
learning takes most recent observations for making a forecast. Next type is double
exponential smoothing where apart from α, another smoothing parameter β is used for change
in trend. There are two types of the trend such as additive trend which gives linear trend
analysis and the other is multiplicative trend gives exponential trend analysis. Finally, triple
exponential smoothing adds seasonality (λ) part from α and β.
There are various studies on time series forecasting that used Holt-Winters method.
Lima et al. (2019) uses Holt-Winters Exponential Smoothing in forecasting economic time
series and compares the accuracy of additive and multiplicative models. Panda (2020)
predicts the spreading of COVID-19 from Indian and its states using ARIMA and Holt-
Winters. Kuzmin et al (2017) employed the Holt–Winters Model to forecast and assess the
efficiency of the methods used to plan a firm’s sales in the upmarket sector. Ribeiro (2019)
evaluate both Holt-Winter’s additive and multiplicative time series model to forecast the
Brazilian natural gas production.

Methodology

Data sources

This study utilizes secondary data obtained from the website of Philippine Statistics
Authority or PSA. Specifically, the quarterly production of abaca fiber in Eastern Visayas
region covered a period of 7 years starting from 2014 to 2020, (see Table 1).

Table 1. Quarterly abaca fiber production in Eastern Visayas (2014-2020)

Quarter
Year
Q1 Q2 Q3 Q4
2014 2856.65 3977.72 3317.62 2697.26
2015 2825.97 3856.02 3326.67 2743.93
2016 2516.62 3810.31 3311.15 2854.53
2017 2572.09 3780.17 3267.92 3050.89
2018 2575.98 3728.62 3227.52 3017.21
2019 2625.73 3695.56 3275.68 3016.46
2020 2244.27 3574.54 3284.62 3032.95
Source: Philippine Statistics Authority.
 20 M.C.F. Pleños

Data analysis

Using the data, future values will be generated using Holt-Winters method. This
technique produces exponentially smoothed values for the level of the forecast, the trend, and
the seasonal adjustment to the forecast. The forecast will depend on the following three
components. These three smoothing equations are:
Level: the average value in the series.
Additive HW Method Multiplicative HW Method
! = "(#! − %!−% ) + (1 − ")( !−1 + &!−1 ) ! = ",-! /%!−% . + (1 − ")( !−1 + &!−1 )
0≤"≤1 0≤"≤1

Trend: the increasing or decreasing value in the series.

Additive HW Method Multiplicative HW Method

&! = 2 ( ! − !−1 ) + (1 − 2)&!−1 &! = 2 ( ! − !−1 ) + (1 − 2)&!−1

0≤2≤1 0≤2≤1

Seasonality: the repeating the short-term cycle in the series.

Multiplicative HW Method
Additive HW Method
%3 = 4(-3 − 3 ) + (1 − 4)%356 7
%3 = 4 8 (1 − 4)%356 0 ≤ 4 ≤ 1
98
0≤4≤1
Where %3 is the seasonality, 3 is the series level, &3 is the trend component, -3 is the observed
value and ", 2, and 4 are the smoothing factor of level, trend and seasonality respectively.
Holt-Winters methods describe two distinct models that depend on the behavior of the
initial series which is based on the analysis of linear trend and seasonality component
(Pelegrini & Fogliatto, 2002). Holt-Winters exponential smoothing model can either be a
multiplicative or additive model. The choice of the model depends on the seasonal component
of the series (Mgale, 2021):
a) In Holt-Winters additive method, seasonal pattern of a series has a constant amplitude.
Seasonal changes in the data stay roughly the same over time and don’t fluctuate in
relation to the overall data. This method is best for data with trend and seasonality that
does not increase over time (Hyndman et al., 2021)
The HW additve model equation is given by:
#3 = :3 + ;3 + <3
b) In Holt-Winters multiplicative method, multiplies the trended forecast by the seasonality.
This method is best for data with trend and seasonality that increases over time. It results
in a curved forecast that reproduces the seasonal changes in the data (Winters, 1960).
The HW multiplicative model equation is given by
#3 = :3 × ;3 × <3
Where :3 represents the trend, ;3 is the seasonal component, and < is the error term
The following procedures were performed in extrapolating quarterly production of abaca
fiber in Eastern Visayas region:
Time Series Forecasting Using Holt-Winters Exponential Smoothing: Application to Abaca … 21

1. Decomposition: This method was applied to separate the seasonality, trend, and error
component of a seasonal time series. A comparison was done to determine whether an
additive or multiplicative Holt-Winters model will be used in prediction. Several error
measures were employed to evaluate the performance in the selection of the Holt-
Winters Model, including RMSE, MSE, MAPE, and MASE.
2. Fitting the model and prediction: After the model was identified and all of the
parameters were determined, the model was used to forecast future values.
3. Forecasting evaluation: Diagnostic checks were carried out to ensure reliability.
Residual analysis was performed whether a model has adequately captured the
information in the data. The residuals are equal to the difference between the
observations and the corresponding fitted values.
>3 = -3 − -?3
A good forecasting method will yield residuals with the following properties for it to be
declared a fitted model:
1. The residuals are uncorrelated. The Ljung-Box Test indicates the presence of these
correlations. If the p-value > 0.05, there is a 95% chance the residuals are
independent.
2. The residuals have zero mean.
3. The residuals have constant variance.
4. The residuals are normally distributed.
R version 3.02 will be used for data entry and data analysis such fitting the time series,
predicting, and diagnostic testing.

Results and discussion

Preliminary data analysis

Eastern Visayas was the top abaca fiber producer with 3.34 thousand metric tons or 21.3
percent share of the total production in the 3rd quarter of 2021 (PSA, 2021). In 2020, the
volume of abaca produced in the Philippines amounted to approximately 71 thousand metric
tons (Statista Research Department, 2020). With abaca as the strongest fiber in the world, the
Philippines supplies 87.5 % of the world’s requirement for abaca fiber and as such is the
number one supplier worldwide (Philippines General Consulate - Vancouver, Canada, 2020).
Hence abaca industry has a significant contribution to the Philippine economy. Figure 1
illustrates the abaca fiber production of Eastern Visayas region starting from 2014 until 2020.
According to PSA, there is a high and steady fiber production since 2010 until 2013.
However, in 2014 a significant drop in abaca fiber production occurred until 2020. When the
super typhoon Yolanda, also known as Haiyan, devastated the provinces of Eastern Visayas
in November 2013, it wiped off abaca plantations, resulting in a drop in abaca fiber supply
in the region. In addition, the abaca bunchy top virus wreaked havoc on the business, wiping
out thousands of hectares of abaca. Figure 2 reflects that the time series of abaca fiber output
demonstrated a periodic up and down movement around the current level or known as
seasonality, with a peak in the second quarter of the year.
22 M.C.F. Pleños

Fig. 1. Time series of abaca fiber production in Eastern Visayas (2014-2020)

Source: own Authors’ calculations.

Fig. 2. Seasonal time series plot of abaca fiber production in Eastern Visayas (2014-2020)
Source: see table 1.

Decomposition

Seasonality, trend, and error are the three components of a seasonal time series. These
three time series components were decomposed in Figure 3. The original time series (top),
the estimated trend component (second from top), the estimated seasonal component (third
from top), and the estimated irregular component (fourth from top) are all shown in the graph
below.
Time Series Forecasting Using Holt-Winters Exponential Smoothing: Application to Abaca … 23

Fig. 3. Decomposition of time series data on abaca fiber production in Eastern Visayas
(2014-2020)
Source: see table 1.

Holt-Winters Method Analysis: Additive and Multiplicative

Holt-Winters time series model can either be of additive or multiplicative seasonality.

With additive model, the behavior is linear where changes over time are consistently made
by the same amount, like a linear trend. In this situation, the linear seasonality has the same
amplitude and frequency. The work of Rahman et al. (2016) focused on the analysis of
seasonal time series data using additive and multiplicative seasonal model of Holt-Winters
method and forecast the monthly revenue using additive HW as the best model. Also in the
previous research of Nurhamidah et al. (2020), it predicted seasonal time series data using
the Holt-Winter exponential smoothing additive model using the number of passengers
departing at Hasanudin Airport in 2009-2019 as the data.

Table 2. Initialization of level, trend, seasonal and exponential smoothing parameters of the
additive and multiplicative HW model for the abaca fiber production time series data
Parameters Additive Multiplicative
"? 1e-04 0.0268
2@ 1e-04 1e-04
4? 0.8132 0.9592
ÂC 3332.2488 3350.8505
&DC -9.6558 -8.4428
%̂C -220.1674 0.8487
%̂E 148.9338 1.0334
%̂F 623.2483 1.2279
%̂G -552.0148 0.89
Source: Author’s calculation and analysis (2022).
24 M.C.F. Pleños

On the other hand, the multiplicative model has an increasing or decreasing amplitude
and/or frequency over time. The study of Koehler et al. (1999) deals with Forecasting Models
and Prediction Interval for the Multiplicative Holt-Winters Method. Found in Table 2 are the
smoothing constants estimates and the values used for the initialization of both multiplicative
and additive methods.
Table 3 shows the values comparing the performance of two Holt-Winters models. It
should be noted that the lower weights produce a smoother line, and the larger ones a more
accentuated line. Thus, a smaller weights for a noisy data gives values that do not fluctuate
along with noise

Table 3. Forecasting performance evaluation of HW models of abaca fiber production in Ea

stern Visayas
Error Measures Additive Multiplicative
Mean Square Error (MSE) 12.65501 7.720112
Root Mean Square Error (RMSE) 327.4811 275.261
Mean Absolute Percentage Error (MA 6.857743 5.840228
PE)
Mean Absolute Scaled Error (MASE) 0.9293798 0.7780083
Source: Author’s calculation and analysis (2022).

The study of Gundalia and Dholakia (2012) examines the performance of the Holt-
Winters method in predicting the maximum and minimum daily temperature time series using
the three standard error measures MSE, MAPE and MAD. In this section, the comparison of
two models, additive and multiplicative, using MSE, RMSE, MAPE AND MASE as error
measures.
Figures 4 and 5 illustrates the plots of additive and multiplicative seasonality and the
variation between two plots are visible. Based on the four measures of evaluation (see Table
3), multiplicative method has a better performance than additive method since it has a lower
MSE, RSME, MAPE, and MASE. This is in parallel to the study of Lima et al. (2019) as it
compares the accuracy of additive and multiplicative Holt-Winters method using e commerce
retail sales.
Hence, in predicting abaca fiber production Holt-Winters model with multiplicative
seasonality was used. This model was utilized to estimate abaca fiber production from 2021-
2023.
Time Series Forecasting Using Holt-Winters Exponential Smoothing: Application to Abaca … 25

Fig. 4. Forecasts on abaca fiber production using Holt-Winters additive method

Source: see table 1.

Fig. 5. Forecasts on abaca fiber production using Holt-Winters multiplicative method

Source: see table 1.

Predictions

By Holt-Winters multiplicative model, forecast values of abaca fiber production was

generated. Table 4 presents the forecasts values of abaca fiber production in Eastern Visayas
for the next three years or 12 quarters, starting 2021 until 2023. As plotted in Figure 6, we
notice that during the 1st quarter (January to March) production of abaca fiber is quite low.
During the 2nd quarter (April to June), abaca fiber production is at its peak while there is a
slight reduction in the 3rd quarter (July to September) and continues to reduce during the 4th
quarter. Figure 7 confirms the seasonality of the time series data using seasonal plots.
26 M.C.F. Pleños

Table 4. Forecasts values on abaca fiber production in Eastern Visayas Region (2021-2023)
Year Q1 Q2 Q3 Q4
2021 2207.241 3537.511 3247.591 2995.921
2022 2170 3500 3210 2958
2023 2133 3463 3173 2921
Source: Author’s calculation and analysis (2022).

Fig. 6. Forecasts on abaca fiber production in Eastern Visayas Region (2021-2023)

Source: see table 1.

Fig. 7. Seasonal plot of the forecasted values (2021-2023)

Source: see table 1.
Time Series Forecasting Using Holt-Winters Exponential Smoothing: Application to Abaca … 27

Forecast Evaluation

We need to check whether the model has appropriately captured the information in the
data. To do this, residuals was verified if they are uncorrelated for each time point, if it has a
mean of zero, a constant variance and is normally distributed. By virtual inspection (see
Figure 8), the plot shows that the distribution of forecast errors is roughly centered on zero,
has a constant variance and is more or less normally distributed. By Ljung-Box, the presence
of correlations in residuals was checked. Based on the results of the test, the p-value is less
than 5% (X-squared =18.15, df = 20, p-value = 0.5776) and implies that there is a 95% chance
the residuals are independent. The Ljung-Box test showed that there is little evidence of non-
zero autocorrelations in the in-sample forecast errors, and the distribution of forecast errors
seems to be normally distributed with mean zero

Fig. 8. Forecast errors for abaca fiber production

Source: see table 1.

Conclusions and recommendations

After both multiplicative and additive models were evaluated and compared using
several error measures, we may infer that the multiplicative method performed well and was
utilized in subsequent production predictions. Based on the predicted values, a major decline
in abaca fiber production happens after the second quarter throughout the year. Overall, there
is a downward trend of abaca fiber production. As a result of the forecasted values, abaca
farm rehabilitation is highly suggested in order to boost abaca fiber production in the region.
Government subsidies, specifically on plantlets and other abaca farming inputs, are
recommended. Extension services must also be made available to abaca farmers, particularly
in the fight against the abaca bunchy top virus, and proper monitoring must be done regularly
to the farmers. It is highly suggested to conduct needs assessment survey among the abaca
farmers in order to understand their challenges and identify their specific needs related to
abaca farming. Added to this is the need to conduct baseline survey on abaca farm households
to assess the impact of projects implemented (e.g. subsidies) especially on their incomes.
28 M.C.F. Pleños

Acknowledgements

No grants were received to support this research and publication.

Bibliography
Australian Bureau of Statistics (2022). Retrieved from
https://www.abs.gov.au/websitedbs/d3310114.nsf/home/time+series+analysis:+the+basics.
Alharbi, N. (2021). Forecasting the COVID-19 Pandemic in Saudi Arabia Using a Modified Singular Spectrum
Analysis Approach: Model Development and Data Analysis.
Analytics Vidhya (2022). Retrieved from https://www.analyticsvidhya.com/blog/2021/08/holt-winters-method-for-
time-series-analysis/ Copyright 2013.
Food and Agriculture Organization (2021). Retrieved from https://www.fao.org/economic/futurefibres/
fibers/abaca0/en/.
Far Eastern Agriculture (2021). Retrieved from https://www.fareasternagriculture.com/crops/agriculture/abaca-the-
philippine-fiber.
Gecili, E., Ziady, A., Szczesniak, R.D. (2021). Forecasting COVID-19 confirmed cases, deaths, recoveries:
Revisiting established time series modeling through novel application for the USA and Italy.
Goodwin, P. (2010). The Holt-Winters Approach to Exponential Smoothing: 50 Years Old and Going Strong.
Retrieved from www.forecasters.org/foresight.
Gundalia, M., Dholakia M.B. (2012). Prediction of maximum/minimum temperatures using Holt-Winters Method
with Excel Spread Sheet for Junagadh Region. International Journal of Engineering Research & Technology,
1(6).
Hyndman, R.J., Athanasopoulos, G. (2021). Holt-Winters’ seasonal method. Forecasting: principals and practice,
3rd Ed. Retrieved from https://otexts.com/fpp3/holt-winters.html.
Koehler, A.B., Snyder, R.D., Keith, J. (1999). Forecasting Models and Prediction Intervals for the Multiplicative
Holt-Winters Method. Department of Econometrics and Business Statistics Monash University
Kuzmin, A.G., Bykov, V.M., Kazaryan, M.A., Danko, T.P., Sekerin, V.D. (2017). Employing the Holt–Winters
Model to Forecast and Assess the Efficiency of the Methods Used to Plan a Firm’s Sales in the Upmarket
Sector. International Journal of Economic Research Vol 14 (7).
Lima, S., Goncalves, M., Costa, M. (2019). Time Series forecasting using Holt-Winters Exponential Smoothing: An
Application to Economic Data. https;//doi.org/10.1063/1.5137999.
Liu, Z., Guo, W. (2020). Government Responses Matter: Predicting COVID-19 cases in US using an empirical
Bayesian time series framework.
Mgale, Y.J., Yan, Y.X., Timothy, S. (2021). A Comparative Study of ARIMA and Holt-Winters Exponential
Smoothing Models for Rice Price Forecasting in Tanzania. Open Access Library Journal, 8, 1-9. DOI:
10.4236/oalib.1107381.
Nusyirwan, N., Ahmad Faisol, F. (2020). Forecasting Seasonal Time Series Data using The Holt-Winters
Exponential Smoothing Method of Additive Models. Jurnal Matematika Integratif 16(2), 151-157.
doi:10.24198/jmi.v16.n2.29293.151-157
Panda, M. (2020). Application of ARIMA and Holt-Winters forecasting model to predict the spreading of COVID
19 for India and its states. https://doi.org/10.1101/2020.07.14.20153908.
Philippines General Consulate - Vancouver, Canada (2020). Retrieved from
https://www.vancouverpcg.org/resources-listofrestaurants/yamang-pinoy/abaca/the-philippine-
abacaindustry.
Philippine Statistics Authority (2021). Retrieved from https://psa.gov.ph/non-food.
Pelegrini, F.R., Fogliatto, F. (2002). Steps for implementation of demand forecasting systems-Techniques and case
study. Magazine Produto & Producao, 11, 43-64.
Rahman, H., Salma, U., Hossain, M., Khan, T.F. (2016) Revenue Forecasting using Holt–Winters Exponential
Smoothing. Research & Reviews: Journal of Statistics, 5(3), 19–25.
Research and Markets. Retrieved from https://www.businesswire.com/news/home/20150112006119/en/Research-
and-Markets-Philippines-Abaca-Fiber-Market-Forecast-and-Opportunities-2019.
Ribeiro, R.C.M., Marques, G.T., Junior, P.C. (2019). Holt-Winters Forecasting for Brazilian Natural Gas
Production. International Educative Research Foundation and Publisher, 7(6), 119-129.
https://doi.org/10.31686/ijier.Vol 7 (6).
Time Series Forecasting Using Holt-Winters Exponential Smoothing: Application to Abaca … 29

Ruekkasaem, L., Sasananan, M. (2018). Forecasting agricultural products prices using time series methods for crop
planning. International Journal of Mechanical Engineering and Technology (IJMET), 9(7), 957-971.
Satrio, C.B.A., Darmawan, W., Nadia, B.U., Hanafiah, N. (2021).Time series analysis and forecasting of coronavirus
disease in Indonesia using ARIMA model and PROPHET. Procedia Comput. Sci., 179, 524-532.
Song, X., Xiao, J., Deng, J., Kang, Q., Qhang, Y., and Xu, J. (2016). Time series analysis of influenza incidence in
Chinese provinces from 2022 to 2011.
SolarWinds Worldwide (2021). Retrieved from https://orangematter.solarwinds.com/2019/12/15/holt-winters-
forecasting-simplified/.
Statista Research Department (2020). Retrieved from https://www.statista.com/statistics/751781/philippines-abaca-
production.
Waller, V., Wilsby, A. (2019). Abaca in the Philippines, an overview of a potential important resource for the
country: Relating the tensile strength of the single fiber to the microfibrilar angle. Retrieved from
http://kth.diva-portal.org/smash/record.jsf?pid=diva2%3A1352495&dswid=8192.
Winters, P.R. (1960). Forecasting sales by exponentially weighted moving averages. Management Science, 6(3),
324–342.

For citation:
Pleños M.C.F. (2022). Time Series Forecasting Using Holt-Winters Exponential Smoothing:
Application to Abaca Fiber Data. Problems of World Agriculture, 22(2), 17-29;
DOI: 10.22630/PRS.2022.22.2.6