=Paper=
{{Paper
|id=Vol-3348/short1
|storemode=property
|title=Simulation of Impact of Migration Caused by the War in Ukraine on COVID-19 Epidemic Process in Moldova
|pdfUrl=https://ceur-ws.org/Vol-3348/short1.pdf
|volume=Vol-3348
|authors=Dmytro Chumachenko,Pavlo Pyrohov,Lyubov Makhota
|dblpUrl=https://dblp.org/rec/conf/profitai/ChumachenkoPM22
}}
==Simulation of Impact of Migration Caused by the War in Ukraine on COVID-19 Epidemic Process in Moldova==
https://ceur-ws.org/Vol-3348/short1.pdf
Simulation of Impact of Migration Caused by the War in
Ukraine on COVID-19 Epidemic Process in Moldova
Dmytro Chumachenko 1, Pavlo Pyrohov 1 and Lyubov Makhota 2
1
National Aerospace University “Kharkiv Aviation Institute”, Chkalow str., 17, Kharkiv 61070, Ukraine
2
State Institution “Kharkiv Regional Center for Disease Control and Prevention of the Ministry of Health of
Ukraine”, Pomirky, Kharkiv 61000, Ukraine
Abstract
The novel coronavirus pandemic has become a global challenge and has shown that health
systems worldwide are unprepared for pandemics of this magnitude. The war in Ukraine,
escalated by Russia on February 24, 2022, brought deaths and a humanitarian catastrophe and
stimulated the spread of COVID-19. Most refugees who evacuated from the war crossed the
border with other countries. At the end of July, almost 550 thousand people crossed the
border with Moldova. This study is devoted to modeling the impact of migration processes
on the dynamics of COVID-19 in Moldova. For this, a machine learning model was built
based on the polynomial regression method. The forecast accuracy a month before the
escalation of the war was from 98.77% to 96.37% for new cases and from 99.8% to 99.75%
for fatal cases. The forecast accuracy for the first month after the escalation of the war was
from 99.96% to 99.34% for new cases and from 99.91% to 99.88% for fatal cases. The high
accuracy of the model, both before the war and with the start of its escalation, suggests that
the migration flows of refugees from Ukraine to Moldova did not affect the dynamics of
COVID-19.
Keywords 1
COVID-19, machine learning, epidemic model, polynomial regression, war
1. Introduction
The COVID-19 pandemic began in December 2019 in Wuhan, China. In the first few months, the
virus spread throughout the planet. The rapid spread of the new coronavirus has led the World Health
Organization to declare a global pandemic of COVID-19. At the end of July 2022, almost 580 million
cases were registered worldwide, almost 6.5 million of which died [1].
In Moldova, the first case of COVID-19 was registered on March 7, 2020. The number of infected
people overgrew, and on March 17, Parliament declared a state of emergency in the country. On
March 18, 2020, the first death was registered. By March 23, the total number of patients with
COVID-19 reached 100 cases; by April 7, it exceeded 1000 cases [2]. At the end of July 2022, more
than 520 thousand cases were registered in Moldova, more than 11.5 thousand of which ended in
death.
Russia's wanton and bloody invasion of Ukraine on February 24, 2022, resulted in numerous
casualties and the destruction of entire cities, including public health infrastructure, leading to a
humanitarian crisis. The escalation of the war has also affected the spread of COVID-19. According
to the International Organization for Migration, the number of internally displaced persons in Ukraine
exceeds 6.2 million [3]. According to the United Nations, the number of refugees from Ukraine in the
countries of the European Union exceeds 6.1 million people [4]. At the end of July 2022, almost 550
2nd International Workshop of IT-professionals on Artificial Intelligence (ProfIT AI 2022), December 2-4, 2022, Łódź, Poland
EMAIL: dichumachenko@gmail.com (D. Chumachenko); pavel.vogorip@gmail.com (P. Pyrohov); labses.centr.kh@ukr.net (L. Makhota)
ORCID: 0000-0003-2623-3294 (D. Chumachenko); 0000-0002-6100-4406 (P. Pyrohov); 0000-0002-1038-5157 (L. Makhota)
©️ 2022 Copyright for this paper by its authors.
Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR Workshop Proceedings (CEUR-WS.org)
thousand people crossed the border with Moldova, and more than 85 thousand people received
refugee status in Moldova.
In the first weeks of the escalation of the war, there were large concentrations of people at border
crossings and in evacuation transport. At the same time, refugees' documents were checked according
to a simplified procedure, and medical documents were not controlled. Mass gatherings, a low level of
vaccination against COVID-19 in Ukraine, and the lack of control of medical records can lead to
outbreaks of infectious diseases in countries hosting refugees.
The COVID-19 pandemic has given impetus to the development of health information tools.
Groups of scientists around the world have developed tools for modeling epidemic processes [5],
analyzing medical data [6], medical diagnostics [7], assessing social factors [8], decision-making for
healthcare organizations [9], and processing medical data [10], etc.
The study aims to build a machine learning model based on the polynomial regression method for
assessing the incidence of COVID-19 in Moldova after the escalation of the Russian war in Ukraine.
The relevance of the study is justified by the fact that the escalation of the Russian war in Ukraine
affects the dynamics of infectious diseases not only in Ukraine, but also in countries that have
received a large number of refugees. Therefore, this study is the first step - testing the hypothesis of
the influence of migration flows on the dynamics of the incidence of COVID-19 in Moldova.
Research is part of a complex, intelligent information system for epidemiological diagnostics, the
concept of which is discussed in [11].
2. Materials and Methods
Regression analysis is a set of methods for studying the influence of one or more independent
variables on the dependent one [12]. Regression analysis reflects only the mathematical dependence
of variables and not a causal relationship. The main tasks for which regression analysis is used are
predicting the value of a dependent variable using independent variables, determining the contribution
of individual independent variables to the variation of the dependent variable, and determining the
degree of determination of the variation of the dependent variable by independent variables.
Polynomial regression is a machine learning algorithm that trains a linear model on non-linear data
[13]. Polynomial regression allows you to use a linear model even if the data is non-linear by adding
additional features to the data.
The polynomial regression model has the following form:
𝑦𝑖 = 𝛽0 + 𝛽1 𝑥𝑖 + 𝛽2 𝑥𝑖2 + ⋯ + 𝛽𝑚 𝑥𝑖𝑚 + 𝜀𝑖 . (1)
A polynomial regression model can be expressed in matrix form:
𝑦1 1 𝑥1 𝑥12 … 𝑥1𝑚 𝛽0 𝜀1 (2)
𝑦2 1 𝑥2 𝑥22 … 𝑥2𝑚 𝛽1 𝜀2
𝑦3 = 1 𝑥3 𝑥32 … 𝑥3𝑚 𝛽
2 + 𝜀3 .
⋮ ⋱ ⋮ ⋮
⋮ ⋮ ⋮ ⋮
[𝑦𝑛 ] [ … 𝑥𝑛 ] [𝛽𝑚 ] [ 𝜀𝑛]
1 𝑥𝑛 𝑥𝑛2 𝑚
Although polynomial regression is a particular case of multiple linear regression, interpreting a
fitted polynomial regression model requires a slightly different perspective. When fitting polynomial
regression, it is often difficult to interpret individual coefficients because the underlying monomials
can be highly correlated. Although orthogonal polynomials can reduce the correlation, it is usually
more informative to consider the fitted regression function as a whole. The point of simultaneous
confidence intervals can then be used to determine the uncertainty of the estimate of the regression
function.
Polynomial regression makes it possible to model non-linear shared data. A polynomial regression
model is more flexible and can model complex relationships. However, models require careful design,
and prior knowledge of the data is required.
To assess the accuracy of the forecast, the mean absolute error was used:
𝑛
1 |𝑌𝑖 − 𝑌̂𝑖 | (3)
𝑀𝐴𝑃𝐸 = ∑ ,
𝑛 𝑌𝑖
𝑖=1
where Yi is the registered incidence for the analyzed period, 𝑌̂𝑖 is the predicted incidence for the
analyzed period, n is the number of periods.
3. Results
The Python programming language was used to implement the study. Data used for modeling
included new cases and deaths provided by the World Health Organization Coronavirus (COVID-19)
Dashboard [1]. To verify the model, a forecast of new cases and deaths from COVID-19 in Moldova
was built for January 25, 2022 - February 23, 2022. Figure 1 shows the forecast for new cases of
COVID-19.
Figure 1: Forecasting of COVID-19 cumulative cases (25.01.22 – 23.02.22)
Table 1 shows the forecast accuracy of the constructed model in Moldova for the period from
January 25, 2022, to February 23, 2022.
Table 1
MAPE (25.01.2022 – 23.02.2022)
Duration of forecast New cases Death cases
7 days 1,22769 % 0,20877 %
10 days 1,67109 % 0,22193 %
20 days 2,62635 % 0,19983 %
30 days 2,30780 % 0,25049 %
Figure 2 shows the forecast of fatal cases of COVID-19 in Moldova from January 25, 2022, to
February 23, 2022.
Figure 2: Forecasting of COVID-19 cumulative death cases (25.01.22 – 23.02.22)
The results of COVID-19 simulations in Moldova a month before the Russian war in Ukraine
showed high accuracy for both short-term and long-term forecasts for 30 days.
To assess the impact of migration processes, a forecast was made of the dynamics of the COVID-
19 epidemic process in Moldova for the first month after the escalation of the Russian war in Ukraine.
Figure 3 shows the forecast of cumulative new cases of COVID-19 in Moldova from February 24,
2022, to March 25, 2022.
Figure 4 shows the forecast of cumulative fatal cases of COVID-19 in Moldova from February 24,
2022, to March 25, 2022.
Table 2 shows the forecast accuracy of the constructed model in Moldova for the period from
February 24, 2022, to March 25, 2022.
Table 2
MAPE (24.02.2022 – 25.02.2022)
Duration of forecast New cases Death cases
7 days 0,04258 % 0,12488 %
10 days 0,03682 % 0,11316 %
20 days 0,13747 % 0,11180 %
30 days 0,65784 % 0,08799 %
Figure 5 shows the forecast of daily new cases of COVID-19 in Moldova from February 24, 2022,
to March 25, 2022.
Figure 6 shows the forecast of daily COVID-19 deaths in Moldova from February 24, 2022, to
March 25, 2022.
Figure 3: Forecasting of COVID-19 cumulative cases (24.02.22 – 25.03.22)
Figure 4: Forecasting of COVID-19 cumulative death cases (24.02.22 – 25.03.22)
Figure 5: Forecasting of COVID-19 daily new cases (24.02.22 – 25.03.22)
Figure 6: Forecasting of COVID-19 daily death cases (24.02.22 – 25.03.22)
Figure 7 shows the forecast deviation of daily new cases of COVID-19 in Moldova from February
24, 2022, to March 25, 2022, from reported cases. Figure 8 shows the deviation of the forecast of
daily fatal cases of COVID-19 in Moldova from February 24, 2022, to March 25, 2022, from
registered cases.
Figure 7: Deviation of COVID-19 daily new cases (24.02.22 – 25.03.22)
Figure 8: Deviation of COVID-19 daily death cases (24.02.22 – 25.03.22)
4. Conclusions
The escalation of the Russian war in Ukraine has stimulated the spread of COVID-19. Identifying
new cases in areas with a destroyed public health infrastructure is challenging. Diagnostics in the
temporarily occupied territories are complex due to the impossibility of conducting laboratory tests.
Treatment is complicated due to the redistribution of medical capacities to help the armed forces and
the affected civilian population. The accumulation of people in bomb shelters and during evacuations
stimulates the spread of the virus. Mentally, people feel the danger of war, not infection with COVID-
19, so anti-epidemic measures are not observed.
Almost 550,000 refugees from Ukraine crossed the border with Moldova. Social distance, mask
regime, and other control measures were not observed during the evacuation. Refugees have not been
tested for COVID-19 vaccination status.
However, the high accuracy of the model built as part of this study shows that the migration flows
of refugees from Ukraine have not been the critical factor of changing in the dynamics of the spread
of COVID-19 in Moldova. It can be concluded that the increase in the incidence associated with the
migration of the population was offset by the natural decrease in the incidence of COVID-19 in
Moldova, which was observed from late February 2022 – early March 2022.
However, to avoid new outbreaks, it is necessary to analyze the medical data of refugees arriving
in third countries. Also, a necessary tool is the prioritization of vaccination against COVID-19 for
refugees who have not completed the entire course of vaccination.
This study is only the first step in the study of the impact of the escalation of the Russian war in
Ukraine on the public health system and, in particular, on changes in the patterns of behavior of the
epidemic process of infectious disease. Therefore, to test the hypothesis about the impact of migration
flows on the dynamics of the COVID-19 epidemic process in Moldova, the simplest statistical
machine learning method, polynomial regression, was chosen.
Future research directions are developing more complex models and methods that will account for
the stochastic nature of the spread of infectious diseases and the heterogeneity of the population. Such
models will make it possible to identify specific factors influencing the epidemic process and evaluate
their information content.
5. Acknowledgements
The study was funded by the National Research Foundation of Ukraine in the frame-work of the
research project 2020.02/0404 on the topic “Development of intelligent technologies for assessing the
epidemic situation to support decision-making within the population biosafety management”
6. References
[1] WHO COVID-19 Dashboard. Geneva: World Health Organization, 2020. Available at:
https://covid19.who.int/ (accessed 25.07.2022).
[2] H. Zach, M. Hanova, M. Letkovicova, Distribution of COVID-19 cases and deaths in Europe
during the first 12 peak weeks of outbreak, Central European Journal of Public Health 19 (1)
(2021): 9-13. doi: 10.21101/cejph.a6394
[3] IOM Ukraine internal displacement report: general population survey – round 6, 23 June 2022,
2022. Available at: https://reliefweb.int/report/ukraine/iom-ukraine-internal-displacement-report-
general-population-survey-round-6-23-june-2022-enuk (accessed 25.07.2022).
[4] Operational Data Portal, Ukraine Refugee Situation. United Nations, 2022. Available at:
https://data.unhcr.org/en/situations/ukraine (accessed 25.07.2022).
[5] E.S. Adamidi, K. Mitsis, K.S. Nikita, Artificial intelligence in clinical care amidst COVID-19
pandemic: a systematic review, Computational and Structural Biotechnology Journal 19 (2021):
2833-2850. doi: 10.1016/j.csbj.2021.05.010.
[6] I. Izonin, et. al., Predictive modeling based on small data in clinical medicine: RBF-based
additive input-doubling method, Mathematical Biosciences and Engineering 18 (3) (2021): 2599-
2613. doi: 10.3934/mbe.2021132
[7] A. Nechyporenko, et. al., Assessment of measurement uncertainty of the uncinated process and
middle nasal concha in spiral computed tomography data, 2019 IEEE International Scientific-
Practical Conference: Problems of Infocommunications Science and Technology, PIC S and T
2019 – Proceedings (2019): 585-588. doi: 10.1109/PICST47496.2019.9061557.
[8] O. Zakharchenko, et. al. Multifaced nature of social media content propagating COVID-19
vaccine hesitancy: Ukrainian case, Procedia Computer Science 198 (2022): 682-687.
doi: 10.1016/j.procs.2021.12.306.
[9] N. Davidich, et. al. Monitoring of urban freight flows distribution considering the human factor,
Sustainable Cities and Society 75 (2021): 103168. doi: 10.1016/j.scs.2021.103168.
[10] O. Skitsan, I. Meniailov, K. Bazilevych, H. Padalko, Evaluation of the informative features of
cardiac studies diagnostic data using the Kullback method, CEUR Workshop Proceedings 2917
(2021): 186-195.
[11] S. Yakovlev, et. al., The concept of developing a decision support system for the epidemic
morbidity control, CEUR Workshop Proceedings 2753 (2020): 265-274.
[12] M. Dou, C. Qin, G. Li, C. Wang, Research on calculation method of free flow discharge based
on artificial neural network and regression analysis, Flow Measurement and Instrumentation 72
(2020) 101707. doi: 10.1016/j.flowmeasinst.2020.101707.
[13] E. Ostertagova, Modelling using Polynomial Regression, Procedia Engineering 48 (2012): 500-
506. doi: 10.1016/j.proeng.2012.09.545