Короткострокове прогнозування основних показників епідемії в Україні на основі моделі сезонних циклів
The authors of this study propose a method of short-term forecasting of time series of the main indicators of the COVID-19 epidemic, which has a pronounced seasonality. This method, which has no direct analogies, provides the decomposition of a general forecasting task into several simpler tasks, su...
Gespeichert in:
| Datum: | 2024 |
|---|---|
| Hauptverfasser: | , , , |
| Format: | Artikel |
| Sprache: | Englisch |
| Veröffentlicht: |
The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute"
2024
|
| Schlagworte: | |
| Online Zugang: | https://journal.iasa.kpi.ua/article/view/322459 |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| Назва журналу: | System research and information technologies |
| Завантажити файл: |  |
Institution
System research and information technologies| _version_ | 1866302996965490688 |
|---|---|
| author | Alyokhin, Alexei Brutman, Anna Grabovoy, Alexandr Shabelnyk, Tetiana |
| author_facet | Alyokhin, Alexei Brutman, Anna Grabovoy, Alexandr Shabelnyk, Tetiana |
| author_sort | Alyokhin, Alexei |
| baseUrl_str | http://journal.iasa.kpi.ua/oai |
| collection | OJS |
| datestamp_date | 2025-02-09T21:55:38Z |
| description | The authors of this study propose a method of short-term forecasting of time series of the main indicators of the COVID-19 epidemic, which has a pronounced seasonality. This method, which has no direct analogies, provides the decomposition of a general forecasting task into several simpler tasks, such as the tasks of building a model of the seasonal cycle of a time series, aggregating the original time series, taking into account the duration of the seasonal cycle, forecasting an aggregated time series, developing an aggregated forecast into a forecast in the original time scale, using the seasonal cycle model. The solution for each task allows the usage of relatively simple methods of mathematical statistics. The article provides a formally rigorous description of all procedures of the method and illustrations of their numerical implementation on the example of a real forecasting task. The use of this method for short-term forecasting of the COVID-19 epidemic development in Ukraine has systematically demonstrated its effectiveness. |
| doi_str_mv | 10.20535/SRIT.2308-8893.2024.4.02 |
| first_indexed | 2025-07-17T10:28:38Z |
| format | Article |
| fulltext |
Publisher IASA at the Igor Sikorsky Kyiv Polytechnic Institute, 2024
32 ISSN 1681–6048 System Research & Information Technologies, 2024, № 4
UDC 616-036.22.001.57:616.9-022:578.834.1
DOI: 10.20535/SRIT.2308-8893.2024.4.02
SHORT-TERM FORECASTING OF THE MAIN INDICATORS
OF THE COVID-19 EPIDEMIC IN UKRAINE BASED
ON THE SEASONAL CYCLE MODEL
A.B. ALYOKHIN, A.B. BRUTMAN, A.N. GRABOVOY, T.V. SHABELNYK
Abstract. The authors of this study propose a method of short-term forecasting of
time series of the main indicators of the COVID-19 epidemic, which has a pro-
nounced seasonality. This method, which has no direct analogies, provides the de-
composition of a general forecasting task into several simpler tasks, such as the tasks
of building a model of the seasonal cycle of a time series, aggregating the original
time series, taking into account the duration of the seasonal cycle, forecasting an ag-
gregated time series, developing an aggregated forecast into a forecast in the original
time scale, using the seasonal cycle model. The solution for each task allows the us-
age of relatively simple methods of mathematical statistics. The article provides a
formally rigorous description of all procedures of the method and illustrations of
their numerical implementation on the example of a real forecasting task. The use of
this method for short-term forecasting of the COVID-19 epidemic development in
Ukraine has systematically demonstrated its effectiveness.
Keywords: COVID-19 epidemic, time series, short-term forecasting, seasonal cycle,
indicators.
INTRODUCTION
Despite the fact that the first works on mathematical epidemiology appeared as
early as the 20s of the last century [1], the COVID-19 pandemic, especially in the
initial phase, created significant difficulties in building high-quality forecasts of
the disease spread based on mathematical models [2–4]. This, along with the
enormous scale of the pandemic, as well as its socio-economic consequences, has
attracted the attention of scientists around the world to the problem of quantitative
forecasting of the development of the COVID-19 epidemic.
Currently, a wide variety of mathematical tools are used to model and fore-
cast the spread of COVID-19, among which the main place is occupied by sys-
temic models of epidemics, including simulation, statistical models and methods
for forecasting time series, methods and models of artificial intelligence (neural
networks and etc.).
LITERATURE ANALYSIS AND PROBLEM STATEMENT
Taking into account the pronounced seasonality of COVID-19 epidemic devel-
opment in various countries of the world, the Box–Jenkins (SARIMA) [5] and
Holt-Winters [6] methods, which take into account seasonal effects, are the most
widely used for statistical modelling of epidemics. SARIMA models are a linear
Short-term forecasting of the main indicators of the COVID-19 epidemic in Ukraine…
Системні дослідження та інформаційні технології, 2024, № 4 33
combination of series and seasonal profile elements, as well as past forecast errors
considering autoregression. The Holt-Winters method uses the technique of triple
exponential smoothing applied to the main components of a statistical time series:
series level, trend, seasonal component. Both models have their own advantages
and disadvantages.
This, in the absence of the only best approach or forecasting model, taking
into account the principle of multiple models generally recognized in forecasting
theory, leaves room for development and determines the relevance of efforts
aimed at developing alternative approaches to forecasting which more accurately
reflect the specifics of the forecasting object, including features of seasonal cy-
clicity, increasing the level of transparency, formalization and simplicity of the
procedures applied.
The paper proposes a new method that decomposes the general problem of
forecasting into a number of simple problems. These are the tasks of building a
model of the seasonal cycle of a time series, aggregating the original time series
taking into account the duration of the seasonal cycle, forecasting the aggregated
time series and transforming the aggregated forecast into a forecast in the original
time scale using the seasonal cycle model.
PURPOSE OF THE STUDY
The aim of the study is to develop a “direct” method for short-term forecasting of
time series of the main indicators of the COVID-19 epidemic based on a seasonal
cycle model.
MATERIALS AND METHODS OF THE STUDY
The study is based on statistical data from the Public Health Center of the Minis-
try of Health of Ukraine [7], as well as indicators derived from them, which to-
gether characterize the spread of the COVID-19 epidemic in Ukraine over the
entire observation period. This statistical data is a multidimensional discrete time
series with a 1-day increment, which includes univariate statistical series of three
basic indicators (increase in infected, deaths, recoveries), five derived daily and
cumulative indicators (increase in active cases; total number of infected, deaths,
recoveries, number of active cases) and three synthetic derived indicators (disease
spread and mortality, epidemic progression).
To determine the availability and duration of a cycle in the statistical series
of reference indicators, autocorrelation methods were used, implemented in the
statistical software IBM SPSS Statistics, STATISTICA and MS Excel 2019. For
aggregating time series, the corresponding series conversion procedures of the
specified application program packages were used.
The building of cycle models, as well as trend models of aggregated time se-
ries of reference indicators, was carried out using generally accepted methods of
analytical alignment of time series and, in particular, curve-fitting methods using
non-linear optimization methods.
To develop and evaluate indicative forecasts of reference indicators, statisti-
cal series were used that included data from the 36th to the 57th week of the
A.B. Alyokhin, A.B. Brutman, A.N. Grabovoy, T.V. Shabelnyk
ISSN 1681–6048 System Research & Information Technologies, 2024, № 4 34
entire observation period. Forecasts were built for a period of 11 weeks, where
the forecast for the second week is an estimate. Statistical series of reference indi-
cators from the 36th to the 55th week were used as training sequences, and the
next two weeks were used as a verification (forecasting) sequence. The accuracy
of forecasts for all forecasted indicators was assessed using the calculated (fore-
casting) data and data from verification sequences using the MAPE accuracy in-
dicator.
MATHEMATICAL MODELS
In the general scheme of the developed method for short-term forecasting of the
COVID-19 epidemic progress, for each one-dimensional statistical series of the
reference indicator, the following procedures are performed:
– identification of the seasonal cycle in the time series and determination of
its duration;
– building of a general model of the seasonal cycle (seasonal profile of the
time series) or a set of models of seasonal cycles of the time series;
– removal of the seasonal component by aggregating the original time series
with a step equal to the cycle duration, and building a trend model of the aggre-
gated series;
– trend forecasting for the forecast period in the aggregated time scale;
– disaggregation of the aggregated forecast using the cycle model (cycle
models), which is the decomposition of the aggregated forecast into a forecast in
the original time scale.
The calculation of the forecast values of all derived indicators of the
COVID-19 epidemic is based on the forecasts of the reference indicators.
For a formal description of the above stated procedures, we introduce the
following notation:
S — original multidimensional statistical time series;
n — dimension of the time series S ;
I — set of indices of the components of the original time series,
},,1{ nI ;
t — discrete moment of time;
sT — length of the time series S ;
FT — duration of the forecasting period;
F — forecasting time series of length FT ;
Let R — n-dimensional vector of COVID-19 epidemic indicators. Then
) ,' , ( * RRRR ,
where *R — vector of epidemic reference indicators of dimension *n ; R — vec-
tor of daily and cumulative derivative indicators of the epidemic of dimension ' n ;
R — vector of synthetic derivative indicators of dimension ".n
In this study, the following assumptions have been made:
} , ,{* RDTCR ,
Short-term forecasting of the main indicators of the COVID-19 epidemic in Ukraine…
Системні дослідження та інформаційні технології, 2024, № 4 35
where RDTC ,, — daily increase of infected, deaths, recovered respectively.
} , , , ,{ ACRDTCACR ,
where AC — daily increase in the number of active cases at time t; , , TC D R —
total number of infected, deaths and recovered; AC — number of active cases.
} , , ,{ PCCTCt IIIRR ,
where tR — infection spread rate; TCI — fatality rate (according to the number
of infected (total cases)); CCI — fatality rate (according to the number of closed
cases); PI — epidemic progress indicator.
The following notations refer to an arbitrary one-dimensional time series
)(iS , where *Ii , where *I — a subset of epidemic reference indices
)|(| ** nI . For simplicity, the index i, where it does not generate ambiguity, will
be omitted below:
CT — length of the seasonal cycle time series )(iS ;
k — number of complete cycles in time series )(iS ;
t — sequence number of the seasonal cycle or, what is the same, a discrete
point in time in the aggregated time scale;
t — sequence number of the observation in the cycle;
)(tSC — t -th statistical time series of length CT (series of seasonal cy-
cle elements t ).
S — aggregated statistical time series of observations;
ST — length of the aggregated time series S in the aggregated time scale
(in units of the number of complete cycles, 'ST k );
FT — duration of the forecast period in the aggregated time scale.
)' (tSCN — t -th normalized seasonal cycle (normalized cycle time series);
CNM — general model of normalized seasonal cycle (seasonal profile of
time series )(iS — model time series of length CT ).
Considering the introduced notation, the formal description of all stages of
forecasting of the reference one-dimensional series )(iS , *Ii , is as follows.
Stage 1. Analysis of the time series cyclicity.
1.1. Determining the availability, duration of CT and the number k of com-
plete cycles in the original time series )(iS .
This stage is implemented by standard methods of autocorrelation analysis.
Stage 2. Building of the normalized cycle model )(iMCN of the time series.
2.1. Normalization of the levels of the seasonal cycle time series ), ( tiSC
for all } , ,1{ , sTtt .
Normalization is carried out according to the formula:
A.B. Alyokhin, A.B. Brutman, A.N. Grabovoy, T.V. Shabelnyk
ISSN 1681–6048 System Research & Information Technologies, 2024, № 4 36
,
)",',(
)",',(
)",',(
"
t
ttir
ttir
ttir
CSt
CS
CN
where )",',( ttirCS — t"-th element of the t'-th cycle of the time series )(iS ;
)",',( ttirCN — t"-th element of the t'-th cycle of the time series )(iS .
2.2. Building of a mathematical model of the normalized cycle )(iMCN
based on ST observations — the set of all cycles }), ({ tiSCN .
The seasonal cycle model )(iMCN is developed on the elements of the nor-
malized time series of cycles )}({ tSCN , the set of which acts as a repetitiveness.
In the study, a degree l polynomial from the sequence number of the cycle ele-
ment was used as a model of the seasonal cycle, where Cl T .
,)()().( 01
1
1 atatatatiM l
l
l
lCN
where } , ,1{ cTt ; A — parameter vector of the degree l polynomial,
),,( 0 laaA .
The parameters A of the )(iMCN model are defined as a solution to an opti-
mization problem of the type:
;}, , 1{, ))",',()",((min 2
,
kttttirtir CNCMt
ttA
;")()()",( 01
1
1 tatatatatir l
l
l
lCM
,"1)",(0 ttirCM
)",( tirCM — t"-th model cycle element )(iMCN ; t — significance coefficient of
the t’-th cycle.
In the model, the weight coefficients tt }{ are set using the logistic func-
tion of the cycle t sequence number
Stage 3. Building a model of the aggregated series )(iS .
3.1. Aggregation of the original one-dimensional time series )(iS — for-
mation of an aggregated time series )(iS of length ST .
The aggregation of the series )(iS is carried out with a step equal to the
duration of the cycle CT , using the operation of summing as the aggregation
operation.
3.2. Building a trend model )(iMT of the aggregated time series )(iS .
It is carried out by an arbitrary method of analytical alignment of time series.
Stage 4. Forecasting of the initial time series )(iS .
4.1. Development of an aggregated forecast )(iF for the forecast time FT .
It is carried out by extrapolating the trend using the )(iMT trend model for
the forecast period FT in the aggregated time scale.
4.2. Formation of the forecasting time series )(iF based on the aggregated
forecast )(iF .
Short-term forecasting of the main indicators of the COVID-19 epidemic in Ukraine…
Системні дослідження та інформаційні технології, 2024, № 4 37
At this stage, deconvolution of each element ) ,( tirF of the forecasting series
)(iF is carried out using the )(iMCN cycle model, as a result, each element is
replaced by the time series of the corresponding cycle t’ in the original time series
scale, the elements of which are determined by the following formula:
1, ),( '
C
FF T
t
irtir };, , 1{, , kTttT
T
t
tiM FC
C
CN
where ][a — an integer part of the number a.
As a result of this stage, the aggregated forecast series )(iF in the aggre-
gated time scale is disaggregated into the forecast series )(iF in the original
time scale.
After forecasting of all COVID-19 epidemic reference indicators, the fore-
cast values of the epidemic derived indicators are calculated in accordance with
the formulas below.
Stage 5. Calculation of the main derivatives (synthetic) indicators of the
COVID-19 epidemic
The daily increase in active cases AC at each time t is derived from the in-
crease in infected TC , deaths D and recovered R and is calculated by the
following formula:
tttt RDTCAC t
Cumulative indicators ACRDTC ,,, are calculated using the following
formulas:
ttt PPP 1 tACRDTCPP ,),,,{, .
The fatality rate is calculated using the formula:
tTC
DI
t
t
TC .
and the fatality rate — according to the formula:
tRD
DI
tt
t
CC )(
As one of the meaningful tools for monitoring and analysing the develop-
ment of the COVID-19 epidemic, the authors propose an PI progress indicator,
which is the ratio of fatality rates:
. )(
)(
)()( tTC
RD
tI
tItI
t
tt
TC
CC
P
А physical meaning of this indicator is transparent, reflected in its name and
indicates the traversed path in % of the epidemic progress from the moment of its
emergence to the moment of completion.
Not least informative is the statistical analogue of the reproduction coeffi-
cient 0R — the infection spread coefficient tR , which is proposed to be calcu-
lated taking into account the seasonality of the time series:
. /
121
tRRR m
Tt
Ttm
l
t
Ttl
t
c
cc
A.B. Alyokhin, A.B. Brutman, A.N. Grabovoy, T.V. Shabelnyk
ISSN 1681–6048 System Research & Information Technologies, 2024, № 4 38
RESULTS AND DISCUSSION
Let us illustrate the implementation of each of the above mentioned stages, using
the example of short-term forecasting of the development of the COVID-19 epi-
demic in Ukraine.
Stage 1. Analysis of the time series cyclicity of the reference indicator.
In the study, to assess the availability and duration of seasonality in these series,
the FORECAST.ETS.SEASONALITY function contained in MS Excel 2019 was
used, which made it possible to detect a seasonal cycle equal to 7 days (see Fig. 1).
The proposed method will be further illustrated for the time series of the in-
dicator of the daily increase in infected (TC).
Stage 2. Building a seasonal cycle model.
As a model of the weekly seasonal profile, a degree 4 polynomial was used,
the parameters of which are determined by solving the corresponding nonlinear
optimization problem. In doing so, the weight coefficients of the actual weekly
cycles were set using the logistic function of the serial number of the cycle.
Stage 3. Building a trend model of an aggregated time series.
In this study, a polynomial no higher than a degree 6 is used as a model of
the aggregated series for illustrative purpose (Fig. 2).
Stage 4. Forecasting the initial time series.
Extrapolation of the aggregated time series is carried out by substituting the
number of the forecast period into the equation of the )(xfy (see Fig. 2), and
DTC dD
dD
Fig. 1. Daily increase in Infected (TC), Deaths (D) and Recovered (R) from COVID-19
in Ukraine (normalized data)
2
1
y = 0.889x6 + 6.0452x5 – 164.44x4 +2329.6 x3 – 17607 x2 + 56337 x + 33674
R2 = 0.9848
Short-term forecasting of the main indicators of the COVID-19 epidemic in Ukraine…
Системні дослідження та інформаційні технології, 2024, № 4 39
disaggregation of the forecast — by deconvolution of the obtained values using
the cycle model.
The results of these procedures for all reference indicators TC, D and R,
as well as the daily increase in active cases AC are shown in Fig. 3.
Step 5. Calculation of derived indicators of the COVID-19 epidemic.
The forecast values of derived indicators are calculated in accordance with the
formulas described above. Their correspondence to the actual data is shown in Fig. 4.
Evaluation of forecast accuracy. In order to evaluate the performance of
the proposed method, the accuracy of forecasts has been assessed for the main
indicators of the COVID-19 epidemic in Ukraine using the mean absolute per-
centage error MAPE, as well as a comparative analysis of the accuracy of this
Dayly increase in Deaths
Daily increase in Recovered
Daily increase in Infected
Dayly increase in Cases
dT
C
D
dR
dA
C
1
1
1
1
2
2
2
2
1 — Forecast, 2 — Actual 1 — Forecast, 2 — Actual
1 — Forecast, 2 — Actual 1 — Forecast, 2 — Actual
Fig. 3. Actual and forecast values of daily indicators of the COVID-19 epidemic in Ukraine
1
1
1
2
2
2
1
2
1 — Forecast, 2 — Actual 1 — Forecast, 2 — Actual
1 — Forecast, 2 — Actual 1 — Forecast, 2 — Actual
Total Cases Total Deaths
Total Recovered Total Active Cases
R
T
C
A
C
D
Fig. 4. Actual and forecast values of cumulative indicators of the COVID-19 epidemic
in Ukraine
A.B. Alyokhin, A.B. Brutman, A.N. Grabovoy, T.V. Shabelnyk
ISSN 1681–6048 System Research & Information Technologies, 2024, № 4 40
method with the exponential smoothing methods and SARIMA mentioned at the
beginning of the paper (see Table).
As follows from the data in Table, the estimation results confirm the fact,
well known in statistical forecasting of time series, that different methods deal
with different situations in different ways, and this fact does not allow us to give
preference to one of them in general.
Estimates (MAPE) of the forecast accuracy of the main indicators COVID-19
epidemic in Ukraine, %
Weekly forecast 2-week forecast
N Indicator
M1 M2 M3 M1 M2 M3
1 TC 0.56 0.17 0.14 0.60 0.30 0.10
2 D 0.66 1.01 0.74 1.27 1.17 0.92
3 R 0.21 0.48 0.31 0.22 1.13 0.65
4 AC 1.91 2.00 2.00 2.76 3.22 2.88
5 TC 12.30 19.28 11.19 10.48 12.83 8.59
6 D 14.51 10.66 13.07 21.03 8.44 10.02
7 R 34.28 35.27 39.06 21.20 36.06 29.90
8 AC 27.70 16.41 17.19 28.85 34.97 31.14
9 ITC 0.16 1.02 0.88 0.70 0.96 1.00
10 ICC 0.45 0.52 0.42 1.26 0.55 0.35
11 IP 0.34 0.50 0.46 0.57 0.92 0.74
where 1M — MAPE of proposed method; 2M — MAPE of exponential smooth-
ing; 3M — MAPE of SARIMA.
As follows from the data in Table 1, the estimation results confirm the fact,
well known in statistical forecasting of time series, that different methods deal
with different situations in different ways, and this fact does not allow us to give
preference to one of them in general.
It should also be noted that the procedures that implement these methods in
the statistical software IBM SPSS Statistics provide for the generation of a series
of models and the selection of the best among them. In assessing the accuracy of
the method proposed in this study base forecast models were used, and the poten-
tial for improvement of forecast models, which is inherent in the method, is not
utilized. The task of realizing this potential to improve the efficiency of the meth-
od is the subject of further research.
The corresponding calculations were carried out on a systematic basis within
the framework of the activities of the “Working Group on Mathematical Model-
ling of Problems Related to the SARS-CoV-2 Coronavirus Epidemic in Ukraine”
of the National Academy of Sciences of Ukraine during 2020 and confirm the
method effectiveness proposed by the authors of the study.
CONCLUSIONS
The time series of daily growth indicators of the COVID-19 epidemic have a pro-
nounced seasonal pattern.
The method proposed by the authors makes full use of this circumstance and
implements the idea of decomposing the general task of developing short-term
Short-term forecasting of the main indicators of the COVID-19 epidemic in Ukraine…
Системні дослідження та інформаційні технології, 2024, № 4 41
forecasts of the main indicators of the COVID-19 epidemic into a number of
particular subtasks, for which simple methods of mathematical statistics are
applicable. In particular, the “direct” method of identifying the seasonal cycle
makes it possible to use quite simple mathematical models of an arbitrary form to
describe the seasonal profile of a time series. Aggregation of the original time
series with a step equal to the seasonal cycle duration enables to eliminate the
seasonal component without distorting the information, and the problem of
forecasting the initial time series can be transformed into the simpler problem of
extrapolating the trend model of the aggregated time series with subsequent
deconvolution (using the cycle model) of its forecasting values in daily values of
the corresponding indicators.
Practical use of this method for developing short-term forecasts of the
COVID-19 epidemic progress in Ukraine on a systematic basis has demonstrated
quite satisfactory accuracy of forecasts (MAPE estimates).
REFERENCES
1. W.O. Kermack, A.G. McKendrick, “A Contribution to the Mathematical Theory of
Epidemics,” Proc. Roy. Soc. Lond. A., no. 115, pp. 700–721, 1972.
2. S. Borenstein, C.K. Johnson, “Modeling coronavirus: ‘Uncertainty is the only cer-
tainty,’” AP News, April 8, 2020. Accessed on: February 10, 2022. [Online]. Avail-
able: https://apnews.com/article/public-health-health-us-news-ap-top-news-virus-
outbreak-88866498ff5c908e5f28f7b5b5e5b695
3. V. Romanenko, Y. Milyavsky, “Combined control of multirate impulse processes in
a cognitive map of COVID-19 morbidity,” System Research and Information Tech-
nologies, no. 3, pp. 46–56, 2022. doi: https://doi.org/10.20535/SRIT.2308-
8893.2022.3.03
4. L Guryanova, R. Yatsenko, N. Dubrovina, V. Babenko, and V. Gvozditskyi, “Ma-
chine learning methods and models, predictive analytics and applications: develop-
ment trends in the post-crisis syndrome caused by COVID-19,” CEUR Workshop
Proceedings, 2927, рр. 1–7, 2021.
5. G.E.P. Box, G.M. Jenkins, G.C. Reinsel, and G.M. Ljung, Time Series Analysis:
Forecasting and Control. New Jersey: John Wiley & Sons, Inc., 2015, 712 p.
6. P.R. Winters, “Forecasting sales by exponentially weighted moving averages” Man-
agement Science, no. 6, pp. 324–342, 1960.
7. “Koronavirusna infektsiia COVID-19 [Corona virus infection COVID-19],” Tsentr
hromadskoho zdorovia MOZ Ukrainy. Available: https://phc.org.ua/kontrol-
zakhvoryuvan/inshi-infekciyni-zakhvoryuvannya/koronavirusna-infekciya-covid-19
8. “Prohnoz rozvytku epidemii COVID-19 v Ukraini na period 7–13 hrudnia 2020 roku
vid uchenykh Instytutu problem rynku ta ekonomiko-ekolohichnykh doslidzhen
NAN Ukrainy [Forecast of the development of the COVID-19 epidemic in Ukraine
for the period of December 7–13, 2020 by scientists of the Institute of Market
Problems and Economic and Environmental Research of the National Academy of
Sciences of Ukraine],” Natsionalna akademiia nauk Ukrainy, December 7, 2020. Avail-
able: https://www.nas.gov.ua/ UA/Messages/Pages/View.aspx?MessageID=7237
Received 30.08.2023
INFORMATION ON THE ARTICLE
Alexei B. Alyokhin, ORCID: 0000-0001-5209-8036, Mariupol State University, Ukraine,
e-mail: aba99@ukr.net
A.B. Alyokhin, A.B. Brutman, A.N. Grabovoy, T.V. Shabelnyk
ISSN 1681–6048 System Research & Information Technologies, 2024, № 4 42
Anna B. Brutman, ORCID: 0000-0002-7774-5356, National University “Zaporizhzhia
Polytechnic”, Ukraine, e-mail: a_brutman@yahoo.com
Alexandr N. Grabovoy, ORCID: 0000-0001-5705-9909, Bogomolets National Medical
University, Ukraine, e-mail: angrabovoy@gmail.com
Tetiana V. Shabelnyk, ORCID: 0000-0001-9798-391Х, Simon Kuznets Kharkiv National
University of Economics, Ukraine, e-mail: tanya.shabelnik17@gmail.com
КОРОТКОСТРОКОВЕ ПРОГНОЗУВАННЯ ОСНОВНИХ ПОКАЗНИКІВ
ЕПІДЕМІЇ В УКРАЇНІ НА ОСНОВІ МОДЕЛІ СЕЗОННИХ ЦИКЛІВ /
О.Б. Альохін, А.Б. Брутман, О.М. Грабовий, Т.В. Шабельник
Анотація. Запропоновано метод короткострокового прогнозування часових
рядів основних показників епідемії COVID-19, яким притаманна виражена се-
зонність. Зазначений метод, що не має прямих аналогів, передбачає декомпо-
зицію загального завдання прогнозування на ряд більш простих завдань, таких
як побудова моделі сезонного циклу часового ряду, агрегування вихідного ча-
сового ряду з урахуванням тривалості сезонного циклу, прогнозування агрего-
ваного часового ряду, розгортання агрегованого прогнозу в прогноз у вихідній
часовій шкалі за допомогою моделі сезонного циклу, вирішення кожного з
яких допускає застосування відносно простих методів математичної статисти-
ки. Наведено формально строге описання всіх процедур методу та ілюстрації
їх числової реалізації на прикладі реального завдання прогнозування. Застосу-
вання зазначеного методу для розроблення короткострокових прогнозів розвитку
епідемії COVID-19 в Україні на систематичній основі продемонструвало його
ефективність.
Ключові слова: епідемія COVID-19, часові ряди, короткострокове прогнозу-
вання.
|
| id | journaliasakpiua-article-322459 |
| institution | System research and information technologies |
| keywords_txt_mv | keywords |
| language | English |
| last_indexed | 2025-07-17T10:28:38Z |
| publishDate | 2024 |
| publisher | The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute" |
| record_format | ojs |
| resource_txt_mv | journaliasakpiua/87/4b7e1c49cd2712e216f85f51cfef0587.pdf |
| spelling | journaliasakpiua-article-3224592025-02-09T21:55:38Z Short-term forecasting of the main indicators of the COVID-19 epidemic in Ukraine based on the seasonal cycle model Короткострокове прогнозування основних показників епідемії в Україні на основі моделі сезонних циклів Alyokhin, Alexei Brutman, Anna Grabovoy, Alexandr Shabelnyk, Tetiana епідемія COVID-19 часові ряди короткострокове прогнозування COVID-19 epidemic time series short-term forecasting seasonal cycle indicators The authors of this study propose a method of short-term forecasting of time series of the main indicators of the COVID-19 epidemic, which has a pronounced seasonality. This method, which has no direct analogies, provides the decomposition of a general forecasting task into several simpler tasks, such as the tasks of building a model of the seasonal cycle of a time series, aggregating the original time series, taking into account the duration of the seasonal cycle, forecasting an aggregated time series, developing an aggregated forecast into a forecast in the original time scale, using the seasonal cycle model. The solution for each task allows the usage of relatively simple methods of mathematical statistics. The article provides a formally rigorous description of all procedures of the method and illustrations of their numerical implementation on the example of a real forecasting task. The use of this method for short-term forecasting of the COVID-19 epidemic development in Ukraine has systematically demonstrated its effectiveness. Запропоновано метод короткострокового прогнозування часових рядів основних показників епідемії COVID-19, яким притаманна виражена сезонність. Зазначений метод, що не має прямих аналогів, передбачає декомпозицію загального завдання прогнозування на ряд більш простих завдань, таких як побудова моделі сезонного циклу часового ряду, агрегування вихідного часового ряду з урахуванням тривалості сезонного циклу, прогнозування агрегованого часового ряду, розгортання агрегованого прогнозу в прогноз у вихідній часовій шкалі за допомогою моделі сезонного циклу, вирішення кожного з яких допускає застосування відносно простих методів математичної статистики. Наведено формально строге описання всіх процедур методу та ілюстрації їх числової реалізації на прикладі реального завдання прогнозування. Застосування зазначеного методу для розроблення короткострокових прогнозів розвитку епідемії COVID-19 в Україні на систематичній основі продемонструвало його ефективність. The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute" 2024-12-25 Article Article application/pdf https://journal.iasa.kpi.ua/article/view/322459 10.20535/SRIT.2308-8893.2024.4.02 System research and information technologies; No. 4 (2024); 32-42 Системные исследования и информационные технологии; № 4 (2024); 32-42 Системні дослідження та інформаційні технології; № 4 (2024); 32-42 2308-8893 1681-6048 en https://journal.iasa.kpi.ua/article/view/322459/312889 |
| spellingShingle | епідемія COVID-19 часові ряди короткострокове прогнозування Alyokhin, Alexei Brutman, Anna Grabovoy, Alexandr Shabelnyk, Tetiana Короткострокове прогнозування основних показників епідемії в Україні на основі моделі сезонних циклів |
| title | Короткострокове прогнозування основних показників епідемії в Україні на основі моделі сезонних циклів |
| title_alt | Short-term forecasting of the main indicators of the COVID-19 epidemic in Ukraine based on the seasonal cycle model |
| title_full | Короткострокове прогнозування основних показників епідемії в Україні на основі моделі сезонних циклів |
| title_fullStr | Короткострокове прогнозування основних показників епідемії в Україні на основі моделі сезонних циклів |
| title_full_unstemmed | Короткострокове прогнозування основних показників епідемії в Україні на основі моделі сезонних циклів |
| title_short | Короткострокове прогнозування основних показників епідемії в Україні на основі моделі сезонних циклів |
| title_sort | короткострокове прогнозування основних показників епідемії в україні на основі моделі сезонних циклів |
| topic | епідемія COVID-19 часові ряди короткострокове прогнозування |
| topic_facet | епідемія COVID-19 часові ряди короткострокове прогнозування COVID-19 epidemic time series short-term forecasting seasonal cycle indicators |
| url | https://journal.iasa.kpi.ua/article/view/322459 |
| work_keys_str_mv | AT alyokhinalexei shorttermforecastingofthemainindicatorsofthecovid19epidemicinukrainebasedontheseasonalcyclemodel AT brutmananna shorttermforecastingofthemainindicatorsofthecovid19epidemicinukrainebasedontheseasonalcyclemodel AT grabovoyalexandr shorttermforecastingofthemainindicatorsofthecovid19epidemicinukrainebasedontheseasonalcyclemodel AT shabelnyktetiana shorttermforecastingofthemainindicatorsofthecovid19epidemicinukrainebasedontheseasonalcyclemodel AT alyokhinalexei korotkostrokoveprognozuvannâosnovnihpokaznikívepídemíívukraínínaosnovímodelísezonnihciklív AT brutmananna korotkostrokoveprognozuvannâosnovnihpokaznikívepídemíívukraínínaosnovímodelísezonnihciklív AT grabovoyalexandr korotkostrokoveprognozuvannâosnovnihpokaznikívepídemíívukraínínaosnovímodelísezonnihciklív AT shabelnyktetiana korotkostrokoveprognozuvannâosnovnihpokaznikívepídemíívukraínínaosnovímodelísezonnihciklív |