Forecasting Currency in Circulation in Nigeria Using Holt-Winters Exponential Smoothing Method
Forecasting Currency in Circulation in Nigeria Using Holt-Winters Exponential Smoothing Method
Forecasting Currency in Circulation in Nigeria Using Holt-Winters Exponential Smoothing Method
Authors’ contributions
This work was carried out in collaboration among all authors. All authors read and approved the final
manuscript.
Article Information
DOI: 10.9734/SAJSSE/2023/v20i1689
Received: 18/04/2023
Data Article Accepted: 22/06/2023
Published: 04/07/2023
ABSTRACT
Nigeria is a country located in West Africa. It is the most populous country in Africa, with a
population of about 211 million. Nigeria is a federal republic with three tiers of government: federal,
state, and local. Nigeria is a diverse country, with over 250 ethnic groups. This study aims to
investigate the forecasting of currencies in circulation (CIC) in Nigeria using the Holt-Winters
exponential smoothing method (Additive Holt–Winters Model and Multiplicative Holt–Winters
Method). The forecasting data are collected from January 1960 to December 2022. The focus of
the study is primarily to determine the optimal forecasting approach while considering the pertinent
smoothing parameters and determine which of the forecasting method, either the multiplicative
_____________________________________________________________________________________________________
S. Asian J. Soc. Stud. Econ., vol. 20, no. 1, pp. 25-41, 2023
Atoyebi et al.; S. Asian J. Soc. Stud. Econ., vol. 20, no. 1, pp. 25-41, 2023; Article no.SAJSSE.101334
Holt–Winters or additive Holt–Winters is best in forecasting CIC in Nigeria. Furthermore, the study
offers prognostic insights regarding CIC in Nigeria. Three experiments were conducted with
different parameters (α= 0.2, γ= 0.2, δ= 0.5; α= 0.4, γ= 0.3, δ= 0.1; α= 0.1, γ= 0.1, δ= 0.3) that were
randomly generated. The experiment with α= 0.4, γ= 0.3, and δ= 0.1 parameters was the best
model for forecasting the variable using both the Additive Holt-Winters model and the Multiplicative
Holt-Winters method. Based on the comparison of the accuracy measures, this study concluded
that the multiplicative Holt-Winters method outperforms the additive Holt-Winters model in all three
measures: MAPE, MAD, and MSD. The multiplicative Holt-Winters method has a significantly lower
MAPE (5.97894E+00) compared to the additive Holt-Winters model (9.55758E+02). The forecast of
CIC in Nigeria was conducted using the multiplicative Holt-Winters method, and it found that
currency in circulation in Nigeria continues to increase as it shows the upper trend.
26
Atoyebi et al.; S. Asian J. Soc. Stud. Econ., vol. 20, no. 1, pp. 25-41, 2023; Article no.SAJSSE.101334
forecasting model [9]. Various statistical and computer simulation using Particle Swarm
graphical tools can aid in this selection process. Optimization. Results from the experiment
One recommended approach is to create showed that Particle Swarm Optimization is
sequence plots of the time series. Sequence capable of calculating reliable smoothing
plots visually display the values of the data series constants, thereby producing forecasts that
over time, with the horizontal axis representing accurately determine the direction (i.e. fall or
the passage of time and the vertical axis rise) of exchange rates. Moreover, the computed
indicating the value. These plots facilitate the Mean Absolute Deviation (MAD) and Residual
identification of key characteristics, such as Standard Error (RSE) from the exchange rate
trends and seasonality, which are further forecasts based on actual observed data suggest
explored in the report. The presence or absence that PSO may also be utilized to enhance
of these components significantly influences forecasting precision.
model selection and prediction accuracy. If there
is a multilevel time series dataset with Lima, et al. [13] have undertaken a study that
hierarchical structure, where the lower-level data examines the prediction of economic time series
exhibits trend and/or seasonality, Holt-Winters that display pronounced trends and seasonal
Exponential Smoothing can be used at each patterns. The researchers have proposed an
level of the hierarchy within a multilevel modeling approach that uses Holt-Winters Exponential
framework. This combination will allow the Smoothing to enhance the likelihood of capturing
modeling and forecasting of both the within- diverse patterns in the data and, thereby,
group and between-group variations in the data. augment forecasting performance. The study
involves a comparative analysis of the accuracy
The foundation of time-series forecasting lies in of Holt-Winters models (additive and
distinguishing between patterns and random multiplicative) in forecasting and aims to provide
errors within a time series [10]. By identifying novel insights into the methods employed
trends, long-term changes, and seasonality through this approach. The selection of these
resulting from factors like variations in usage and methods is based on their capacity to model
demand, the goal is to separate the underlying trends and seasonal fluctuations that are
pattern from the noise. Several time series inherent in economic data. The models are fitted
forecasting techniques are available, including to time series of e-commerce retail sales in
Moving Averages, Linear Regression with Time, Portugal, and the study evaluates the accuracy
and Exponential Smoothing. Olayiwola [11] of additive and multiplicative output.
developed ARIMA (4,1, 4) model to predict the
road traffic fatality in South Africa. This model is Olayiwola and Atoyebi [14] conducted an
concluded to be the best among the set of investigation on the Box-Jenkins Approach to
models considered duw to its lowest volatility, Fuel Prices and Currency Strength in the
lowest information criterion and highest adjusted International Market. The study used an ARIMA
R-squared values. In this report, the focus is to predict the fuel prices and the strength of the
primarily to determine the optimal forecasting South African rand in the international market,
approach while considering the pertinent based on 35 years of monthly data. The fuel
smoothing parameters and determine which of prices were observed to demonstrate an upward
the forecasting method, either the trend variation. Upon estimating numerous
multiplicativeHolt–Winters or additiveHolt– competing models for each variable, ARIMA (3,
Winters is best in forecasting CIC in Nigeria. 1, 1), ARIMA (3, 1, 1), ARIMA (1, 1, 2) and
Furthermore, this study offers prognostic insights ARIMA (1, 0, 1) were found to be the most
regarding CIC in Nigeria. suitable models for modelling and predicting the
future values of diesel, paraffin, petrol and
2. LITERATURE REVIEW exchange rate (ZAD-USD), respectively. The
corresponding forecasting accuracies were
Ortiz [12] conducted a study on the precision rate 93.4%, 91.7%, 91.5% and 79.3%. Their study
of the Holt-Winters model in forecasting concluded that using these models in predicting
exchange rates by using particle swarm the future values of fuel prices and the strength
optimization. In particular, an exponential of South African currency against the United
smoothing method, through the use of the Holt- States of America dollars can help policymakers
Winters Model, was used for the purpose of and all stakeholders make informed decisions in
forecasting exchange rates. The search for their planning. Olayiwola and Seeletse [15]
optimal smoothing constants was performed via earlier in 2020 noted that, an upward variation
27
Atoyebi et al.; S. Asian J. Soc. Stud. Econ., vol. 20, no. 1, pp. 25-41, 2023; Article no.SAJSSE.101334
(modelled by quadratic rate) of petrol price would inefficiency in capturing certain effects and
continue into the future. seasonality, possibly due to variations in intra-
month and intra-week lags. In response,
In their study on the currency in circulation in the Dheerasinghe proposed an alternative univariate
country, Mwale et al. [16] emphasized the modelling approach that aimed to independently
importance of CIC and identified two indicators to address trends, seasonal patterns, and cycles in
assess its significance: the proportion of CIC to the series, considering the drawbacks of existing
the total money supply (CIC/money supply) and methods. Successfully identifying cyclical
its relationship to the GDP. An increase in the dynamics in the data, Dheerasinghe and her
percentage of CIC in the money supply indicated team incorporated autoregressive and moving
reduced funds available for lending, potentially average (ARMA) terms into their modelling
hindering economic growth by limiting the approach. They employed time and time-squared
amount of money in banks and deposit series to capture the stochastic trend in the data,
institutions. On the other hand, a high level of accounting for both linear and non-linear trends.
CIC suggested increased economic transactions, Based on their findings, all three approaches
which could contribute to inflationary pressures adequately fit the data and captured various
and indicate an economic boom. Analyzing effects and seasonality, displaying satisfactory
annual data from 1965 to 2004, Mwale et al. performance in out-of-sample forecasting for Sri
discovered a seasonal trend in the data Lanka. The model selection process involved
attributed to seasonal agricultural activity. Mwale criteria such as the lowest Akaike and Schwartz
et al. [16] undertook the task of modelling the information criteria, maximization of R-squared,
level of CIC in Nigeria using a traditional and minimization of mean square error of the
multivariate regression model based on the forecasts.
demand for money. Despite incorporating various
variables, such as nominal GDP growth rate, Norat [18] employed a structural time series
interest rates, shadow economy indicators, (STS) technique, following a similar approach to
indicators of small-scale agricultural activities, Dheerasinghe, to forecast the CIC in Nigeria.
electronic transactions, and a dummy variable, The STS model utilized structural equations to
the model's explanatory power was limited, incorporate trends, seasonality, cycles, and other
accounting for only around 60% of the properties of the time series. Norat applied this
fluctuations in the CIC/money supply. The model to analyze CIC data from the United
authors did observe a negative relationship Kingdom, specifically considering volatile periods
between deposit rates and the amount of money such as Christmas and New Year's celebrations,
in circulation, suggesting that higher deposit during the years 2005 to 2006. By comparing the
rates led to reduced money supply due to outcomes with those of an exponential
increased saving behaviour. Additionally, the smoothing model, Norat [18] claimed that the
model indicated significant effects of small-scale STS model demonstrated superior performance
agricultural production activities and the shadow outside the sample. Moreover, Norat [18]
economy. However, the GDP growth rate and a established that smaller out-of-sample forecast
dummy election variable proved to have marginal errors were obtained when the estimates of the
usefulness in predicting the movement of the STS forecast were weighted with CIC demand
CIC/money supply. from participants in the British Notes Circulation
Scheme.
Dheerasinghe [17] emphasized the importance of
estimating the currency in circulation (CIC) for Griffiths and Desmond [19], while primarily
the Sri Lankan Central Bank, highlighting its focusing on forecasting pandemic data during the
significance as a leading indicator of economic Covid-19 pandemic, introduced the SIR model
growth and accounting for over 65% of the total (Vulnerable, Infected, and Removed) as a
reserve funds. While acknowledging the system of initial value problems formulated as
unsuitability of commonly used approaches for ordinary differential equations. Although they
high-frequency data, Dheerasinghe agreed with employed multiplicative Holt-Winters exponential
the Reserve Bank of India's concerns. She also smoothing, the authors acknowledged that this
recognized the limitations of the traditional method is suitable only when trend patterns are
demand for money model, which lacked more the sole influencing factor in the data. They also
frequent income data beyond quarterly periods. noted that the method fails to capture seasonal
Furthermore, she observed that the alternative trends when the data exhibits both trend and
approach of univariate modelling had shown seasonality. To address this limitation, they
28
Atoyebi et al.; S. Asian J. Soc. Stud. Econ., vol. 20, no. 1, pp. 25-41, 2023; Article no.SAJSSE.101334
incorporated an additional parameter into Holt- forecast CIC. Mwale et al. utilized a traditional
Winters exponential smoothing. However, the multivariate regression model that captured
study used the multiplicative methodology of some factors influencing CIC but had limited
Holt-Winters exponential smoothing without explanatory power. Dheerasinghe proposed an
providing a comprehensive evaluation of its alternative univariate modelling approach aiming
effectiveness or comparing it with alternative to address trends, seasonal patterns, and cycles,
models. demonstrating satisfactory performance.
However, it is crucial to consider the limitations
The study conducted by Riazuddin and Khan [20] and areas for improvement in these models.
focused on modelling CIC and advocated for the
effectiveness of the ARIMA model. They argued 3. METHODOLOGY
that incorporating Islamic calendar effects into
the ARIMA model enhanced its performance. By Accurate prediction is crucial for projecting future
including Islamic event dummy variables, they performance and plays a vital role in
claimed to improve the prediction of cash comprehensive planning and management
movement in circulation in Pakistan during the processes. Forecasting techniques can be
estimation sample period from July 1972 to June grouped into four categories: time series
1999. However, the extent of improvement and univariate, multivariate or causal, qualitative or
the reliability of this approach were not technological, and other quantitative methods. In
extensively discussed. The out-of-sample the study conducted by Guizzi et al. [21], which
forecasts exhibited a mean absolute percent focused on modelling temperature behavior in
error of 0.504 percent, which may seem low, but Caserta, Italy, the authors compared exponential
the lack of contextual comparison or smoothing models and autoregressive moving
comprehensive evaluation of other forecasting average models. They employed the Holt-
methods hampers a thorough assessment. Winters method, a statistical technique within the
time series univariate approach. Holt [22]
Riazuddin and Khan [20] , and Norat supported extended the use of exponentially weighted
the effectiveness of their respective approaches moving averages to account for trend and
without thoroughly evaluating the extent of seasonal variations, while Taylor [23] utilized the
improvement and the robustness of these Holt-Winters approach for short-term electricity
models. Balli and Elsamadisy claimed the demand forecasting. DeLurgio [24] noted that the
superiority of the ARIMA model over the Winters method and Fourier series analysis are
univariate regression model but lacked also adaptable techniques that capture the trend,
comprehensive details in their analysis. Griffiths level, and seasonality of time series data. Puah
and Higham introduced the SIR model with et al. [25] used the additive Holt-Winters method
multiplicative Holt-Winters exponential to analyze rainfall patterns in the Langat River
smoothing, yet they did not extensively evaluate Basin in Malaysia. The Holt-Winters exponential
or compare it with alternative approaches. Thus, smoothing approach is widely employed for
further scrutiny and comprehensive evaluation forecasting various time series. This study aims
are necessary to ascertain the reliability and to predict the monthly CIC in Nigeria, fit an
effectiveness of these models in accurately additive Holt-Winters exponential smoothing
forecasting CIC. This current study aims to model, and evaluate the accuracy measures.
investigate other methods of forecasting CIC in
Nigeria. 3.1 Method of Smoothing
In conclusion, the reviewed studies shed some In the field of time series forecasting, it is
light on the forecasting of CIC using different common for data to exhibit trend patterns, which
models. The ARIMA model showed superior can be characterized by consistent increases or
performance compared to conventional money decreases in the data. The Holt-Winters
demand models in CIC forecasting, but it exponential smoothing technique is a forecasting
suffered from limitations due to its reliance on method that uses exponential smoothing based
historical data and lack of a theoretical economic on previous forecasting results. This technique
foundation. Money demand models and incorporates a parameter to effectively handle
univariate approaches, despite their theoretical seasonal patterns in the data [26,27]. It is
promise, failed to accurately capture the impact suitable for predicting time-series data that
of high-frequency data. The studies under review exhibits both seasonal and trend patterns
employed different modelling approaches to simultaneously. Smoothing, as described by Abd
29
Atoyebi et al.; S. Asian J. Soc. Stud. Econ., vol. 20, no. 1, pp. 25-41, 2023; Article no.SAJSSE.101334
Jalil [28], involves calculating the average value Seasonal patterns smoothing (at the time t)
over multiple years to estimate the value of a
specific year. According to Makridakis et al. [29],
the smoothing approach can be categorized into
two types: the exponential smoothing method
and the smoothing method. Exponential So, the p-period forecasting forward is
smoothing is used to forecast data influenced by
seasonal or trend patterns by assigning different
weights to historical data with exponentially
decreasing significance [29]. where (0 < , , δ < 1).
30
Atoyebi et al.; S. Asian J. Soc. Stud. Econ., vol. 20, no. 1, pp. 25-41, 2023; Article no.SAJSSE.101334
Consequently, multiple forecasting models will be proposed for predicting CIC in Nigeria using the
derived with varying parameters in this study. multiplicative Holt-Winters exponential smoothing
Determination of the most suitable model for method and AdditiveHolt–Winters Model. Three
predicting CIC in Nigeria using this method can trials and errors were conducted with different
be achieved by employing the MAPE value. parameters that were randomly generated; the
Specifically, the chosen model that will be results of the random numbers are; α= 0.2, γ=
considered better will be the model with the 0.2, δ= 0.5; α= 0.4, γ= 0.3, δ= 0.1; α= 0.1, γ= 0.1,
smallest MAPE value. The following models are δ= 0.3.
31
Atoyebi et al.; S. Asian J. Soc. Stud. Econ., vol. 20, no. 1, pp. 25-41, 2023; Article no.SAJSSE.101334
Experimental 1:
32
Atoyebi et al.; S. Asian J. Soc. Stud. Econ., vol. 20, no. 1, pp. 25-41, 2023; Article no.SAJSSE.101334
Experimental 2:
33
Atoyebi et al.; S. Asian J. Soc. Stud. Econ., vol. 20, no. 1, pp. 25-41, 2023; Article no.SAJSSE.101334
Experimental 3:
34
Atoyebi et al.; S. Asian J. Soc. Stud. Econ., vol. 20, no. 1, pp. 25-41, 2023; Article no.SAJSSE.101334
Based on these accuracy measures, it revealed parameters associated with level, trend, and
that Experimental 2 has the lowest MAPE seasonal variables (Bermudez, Segura, and
(5.97894E+00), indicating a lower average Vercher, 2006). It is a specialized exponential
percentage error. Experimental 2 also has the smoothing method tailored for handling seasonal
lowest MAD (2.94826E+05), representing a data.
lower mean absolute deviation. Additionally,
Experimental 2 has the lowest MSD 4.2.2 Additive Holt-Winters method
(2.48397E+11), indicating a lower mean squared
deviation. Considering these factors, The additiveHolt–Winters method is presented in
Experimental 2 performs the best across all three the following equations.
accuracy measures. Therefore, Experimental 2,
with α= 0.4, γ= 0.3, and δ= 0.1, is the best model
for forecasting the variable using the
multiplicative Holt-Winters model. Considering Estimate of the level at time T
the Additive Holt-Winters model, based on the
accuracy measures, it can be inferred that
Experimental 2 has the lowest MAPE
(9.55758E+02), indicating a lower average Estimate of the growth rate (or trend) at time T
percentage error. Experimental 2 also has the
lowest MAD (2.99833E+05), representing a
lower mean absolute deviation. Additionally,
Experimental 2 has the lowest MSD Estimate of the seasonal factor at time T
(1.94250E+11), indicating a lower mean squared
deviation. Experimental 2 also performs the best
across all three accuracy measures. Therefore,
Experimental 2, with α= 0.4, γ= 0.3, and δ= 0.1,
is still the best model for forecasting the variable p- Step ahead forecast made at time T
using the Additive Holt-Winters model.
35
Atoyebi et al.; S. Asian J. Soc. Stud. Econ., vol. 20, no. 1, pp. 25-41, 2023; Article no.SAJSSE.101334
Normalize Constant =
2
Adjusted R :
Initial seasonal factors =
36
Atoyebi et al.; S. Asian J. Soc. Stud. Econ., vol. 20, no. 1, pp. 25-41, 2023; Article no.SAJSSE.101334
4.2.6 Additive Holt–Winters model: α= 0.4, γ= compared to the additive Holt-Winters model
0.3, δ= 0.1 (9.55758E+02). Similarly, the multiplicative Holt-
Winters method has lower MAD and MSD
Comparing the accuracy measures, it can be values. Based on these accuracy measures, it
inferred that the multiplicative Holt-Winters can be concluded that the multiplicative Holt-
method outperforms the additive Holt-Winters Winters method is better for forecasting the CIC
model in all three measures: MAPE, MAD, and in Nigeria than the additive Holt-Winters model.
MSD. The multiplicative Holt-Winters method Therefore, the multiplicative Holt-Winters method
has a significantly lower MAPE (5.97894E+00) forecasts the monthly CIC in Nigeria.
37
Atoyebi et al.; S. Asian J. Soc. Stud. Econ., vol. 20, no. 1, pp. 25-41, 2023; Article no.SAJSSE.101334
38
Atoyebi et al.; S. Asian J. Soc. Stud. Econ., vol. 20, no. 1, pp. 25-41, 2023; Article no.SAJSSE.101334
39
Atoyebi et al.; S. Asian J. Soc. Stud. Econ., vol. 20, no. 1, pp. 25-41, 2023; Article no.SAJSSE.101334
Mathematics and Statistics Studies. 25. Puah YJ, Huang YF, Chua KC, Lee TS.
2020;3(5):63-74. River catchment rainfall series analysis
12. Ortiz RRL. The accuracy rate of holt- using additive holt–winters method.
winters model with particle swarm Journal of Earth System Science. 2016;
optimization in forecasting exchange rates 125(2):269–283.
parameters. 2015;8:9. DOI: 10.1007/s12040-016-0661-6
13. Lima S, Gonçalves AM, Costa M. Time 26. Kalekar P. Time series forecasting using
series forecasting using holt-winters holt-winters exponential smoothing; 2004.
exponential smoothing: an application to 27. Pongdatu GAN, Putra YH. Seasonal time
economic data. In AIP Conference series forecasting using SARIMA and holt
Proceedings; 2019. winter's exponential smoothing In IOP
14. Olayiwola MF, Atoyebi SB. Box-jerkins conference series. in Materials Science
approach to fuel prices and currency and Engineering. 2018;407(1):012153.
strength in the international market. 28. Abd Jalil NA, Ahmad MH, Mohamed N.
Journal of Modern Mathematics and Electricity load demand forecasting using
Statistics. 2021;15(1):8-15. exponential smoothing methods. World
15. Olayiwola MF, Seeletse SM. Statistical Applied Sciences Journal. 2013;22(11):
forecasting of petrol price in South Africa. 1540-1543.
Journal of Engineering and Applied 29. Makridakis S, Wheelwright SC, Hyndman
Sciences. 2020a;15(2):602-606. RJ, McGee VE. Forecasting Methods and
16. Mwale M, Msosa C, Sichinga O, Simwaka Application. Hoboken: John Willey & Sons,
K, Chawani R, Palamuleni A. Currency in Inc; 1999.
Circulation in Malawi; 2004. 30. Panda M. Application of ARIMA and holt-
17. Dheerasinghe R. Modeling and forecasting winters forecasting model to predict the
currency in circulation in Sri Lanka. Staff spreading of covid-19 for India and its
Studies. 2006;36:37-72. states. MedRxiv; 2020.
18. Norat MA. Forecasting banknotes. 31. Makatjane N, Moroke N. Comparative
Handbooks, centre for central banking study of holt-winters triple exponential
studies, bank of england bank of Englnad. smoothing and seasonal arima: forecasting
2008;28. short term seasonal car sales in South
19. Griffiths DF, Desmond JH. Numerical Africa. Risk Governance & Control:
methods for ordinary differential equations: Financial Markets & Institutions,Winter.
Initial value problems. London:Springer. 2016;6(1).
2010;5. 32. Djauhari MA, Asrah NM, Li LS, Djakaria I.
20. Riazuddin R, Khan M. Detection and Forecasting model of electricity
forecasting of islamic calendar effects in consumption in Malaysia: A geometric
time series data. State Bank of Pakistan brownian motion approach. Solid State
Research Bulletin. 2005;1(1):25–34. Technology. 2020;63(3):40-46.
21. Guizzi G, Silvestri C, Romano E, Revetria 33. Bezerra AKL, Santos ÉMC. Prediction the
RA. Comparison of forecast models to daily number of confirmed cases of
predict weather parameters. Advances in COVID-19 in Sudan with ARIMA
Energy and Environmental Science and and holt winter exponential
Engineering. 2015:88-96. smoothing International. Journal of
22. Holt C. Forecasting seasonals and trends Development Research. 2020;10(8):
by exponentially weighted moving 39408-39413.
averages. Int. J. Forecasting. 2004;20(1): 34. Kurita J, Sugawara T, Ohkusa Y. Forecast
5–10. of the COVID-19 outbreak and effects of
23. Taylor J. Short-term electricity demand self restraint in going out in Tokyo Japan.
forecasting using double seasonal MedRxiv; 2020.
exponential smoothing. Journal of the 35. Gupta R, Pal SK. Trend analysis and
Operational Research Society. 2003;54(8): forecasting of COVID-19 outbreak in India.
799–805. MedRxiv; 2020.
DOI: 10.1057/palgrave.jors.2601589 36. Jere S, Siyanga M. Forecasting inflation
24. DeLurgio S. Forecasting principles and rate of Zambia using holt's exponential
applications. New York: McGraw Hill. smoothing open. Journal of Statistics.
1998:205. 2016;6(2):363-372.
40
Atoyebi et al.; S. Asian J. Soc. Stud. Econ., vol. 20, no. 1, pp. 25-41, 2023; Article no.SAJSSE.101334
37. Djakaria I. Djalaluddin Gorontalo airport 39. Atoyebi OM, Atoyebi SB. A Meta-analytic
passenger data forecasting with holt's- Review of the Relationship between
winters' exponential smoothing Mathematics Anxiety and the Mathematical
multiplicative event-based method IOP Thinking of Africa Students. Archives of
conference series. in Journal of Physics; Current Research International. 2022;
2019. 22(7), pp.15-28. DOI: 10.9734/ACRI/2022/
38. Sumitra ID, Basri I. Forecasting the v22i7536
number of outpatient patient visits 40. Atoyebi OM, Atoyebi SB. The link between
using the ARIMA, SES and holt- mathematics teaching strategies and
winters methods at XYZ community students' anxiety. Asian Journal of
health center IOP conference series. in Education and Social Studies. 2022; 33(4),
Materials Science and Engineering; pp.48-57.
2020. DOI: 10.9734/AJESS/2022/v33i4716
© 2023 Atoyebi et al.; This is an Open Access article distributed under the terms of the Creative Commons Attribution License
(http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium,
provided the original work is properly cited.
Peer-review history:
The peer review history for this paper can be accessed here:
https://www.sdiarticle5.com/review-history/101334
41