Predictive Models for Decision Support in the COVID-19 Crisis

103 Pages·2021·4.777 MB·English

by  Joao Alexandre Lobo Marques, Francisco Nauber Bernardo Gois, José Xavier-Neto, Simon James Fong

SPRINGER BRIEFS IN
APPLIED SCIENCES AND TECHNOLOGY
Joao Alexandre Lobo Marques
Francisco Nauber Bernardo Gois
José Xavier-Neto
Simon James Fong
Predictive Models
for Decision
Support in
the COVID-19
Crisis
SpringerBriefs in Applied Sciences
and Technology
SpringerBriefs present concise summaries of cutting-edge research and practical
applications across a wide spectrum offields. Featuring compact volumes of 50 to
125 pages, the series covers a range of content from professional to academic.
Typical publications can be:
(cid:129) A timely report of state-of-the art methods
(cid:129) Anintroductiontooramanualfortheapplicationofmathematicalorcomputer
techniques
(cid:129) A bridge between new research results, as published in journal articles
(cid:129) A snapshot of a hot or emerging topic
(cid:129) An in-depth case study
(cid:129) Apresentation ofcore conceptsthatstudents mustunderstand inordertomake
independent contributions
SpringerBriefs are characterized by fast, global electronic dissemination,
standard publishing contracts, standardized manuscript preparation and formatting
guidelines, and expedited production schedules.
On the one hand, SpringerBriefs in Applied Sciences and Technology are
devoted to the publication of fundamentals and applications within the different
classical engineering disciplines as well as in interdisciplinary fields that recently
emerged between these areas. On the other hand, as the boundary separating
fundamental research and applied technology is more and more dissolving, this
series isparticularlyopentotrans-disciplinary topics between fundamentalscience
and engineering.
Indexed by EI-Compendex, SCOPUS and Springerlink.
More information about this series at http://www.springer.com/series/8884
Joao Alexandre Lobo Marques
(cid:129)
Francisco Nauber Bernardo Gois
(cid:129)
é
Jos Xavier-Neto Simon James Fong
(cid:129)
Predictive Models
for Decision Support
in the COVID-19 Crisis
123
JoaoAlexandre LoboMarques Francisco NauberBernardo Gois
Laboratory of Neuroapplications MachineLearning Department
University of Saint Joseph Secretary of Healthof the Government
Macau, Macao of the State ofCeara
Fortaleza, Brazil
JoséXavier-Neto
Government Intelligence Cell Simon James Fong
Secretary of Healthof the Government Department ofComputer
of the State ofCeara andInformation Science
Fortaleza, Brazil University of Macau
Macau, Macao
ISSN 2191-530X ISSN 2191-5318 (electronic)
SpringerBriefs inApplied SciencesandTechnology
ISBN978-3-030-61912-1 ISBN978-3-030-61913-8 (eBook)
https://doi.org/10.1007/978-3-030-61913-8
©TheEditor(s)(ifapplicable)andTheAuthor(s),underexclusivelicensetoSpringerNature
SwitzerlandAG2021
Thisworkissubjecttocopyright.AllrightsaresolelyandexclusivelylicensedbythePublisher,whether
thewholeorpartofthematerialisconcerned,specificallytherightsoftranslation,reprinting,reuseof
illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmissionorinformationstorageandretrieval,electronicadaptation,computersoftware,orbysimilar
ordissimilarmethodologynowknownorhereafterdeveloped.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publicationdoesnotimply,evenintheabsenceofaspecificstatement,thatsuchnamesareexemptfrom
therelevantprotectivelawsandregulationsandthereforefreeforgeneraluse.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, expressed or implied, with respect to the material contained
hereinorforanyerrorsoromissionsthatmayhavebeenmade.Thepublisherremainsneutralwithregard
tojurisdictionalclaimsinpublishedmapsandinstitutionalaffiliations.
ThisSpringerimprintispublishedbytheregisteredcompanySpringerNatureSwitzerlandAG
Theregisteredcompanyaddressis:Gewerbestrasse11,6330Cham,Switzerland
Contents
1 Prediction for Decision Support During the COVID-19
Pandemic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Explanation and Prediction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2.1 Explanatory Approaches—Data Regression
in Epidemiology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Prediction for Decision Support. . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3.1 Scope and Time Spam . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 The Proposed Approach for the Methodology and Results
Presentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.5 Methodology and Scope Definition for the Data Analysis. . . . . . . 7
1.5.1 Data Source and Data Selection . . . . . . . . . . . . . . . . . . . 7
1.5.2 Methodology for Validation and Evaluation. . . . . . . . . . . 10
1.6 Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2 Epidemiology Compartmental Models—SIR, SEIR, and SEIR
with Intervention . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1.1 Literature Review. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2 Materials and Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2.1 SIR Model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2.2 SEIR Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.2.3 SEIR with Intervention Model . . . . . . . . . . . . . . . . . . . . 20
2.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.3.1 SIR Model—Results and Discussion. . . . . . . . . . . . . . . . 21
2.3.2 SEIR Model—Results and Discussion. . . . . . . . . . . . . . . 27
2.3.3 SEIR Model with Intervention . . . . . . . . . . . . . . . . . . . . 29
2.4 Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
v
vi Contents
3 Forecasting COVID-19 Time Series Based on an Autoregressive
Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.2 Materials and Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.2.1 Proposed Methodology for the Data Analysis . . . . . . . . . 43
3.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.3.1 ARIMA Predictions for China . . . . . . . . . . . . . . . . . . . . 43
3.3.2 ARIMA Predictions for the United States of America . . . 45
3.3.3 ARIMA Predictions for Brazil . . . . . . . . . . . . . . . . . . . . 47
3.3.4 ARIMA Predictions for Italy . . . . . . . . . . . . . . . . . . . . . 49
3.3.5 ARIMA Predictions for Singapore . . . . . . . . . . . . . . . . . 49
3.3.6 Model Performance Comparison Between Countries . . . . 52
3.4 Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4 Nonlinear Prediction for the COVID-19 Data Based on Quadratic
Kalman Filtering. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.2 Materials and Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.2.1 Kalman Filter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.2.2 State-Space Derivation . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.2.3 Quadratic Kalman Filter . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.2.4 Proposed Methodology for the Data Analysis . . . . . . . . . 58
4.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.3.1 Kalman Filter Predictions for China . . . . . . . . . . . . . . . . 58
4.3.2 Kalman Filter Predictions for the United States . . . . . . . . 59
4.3.3 Kalman Filter Predictions for Brazil . . . . . . . . . . . . . . . . 59
4.3.4 Kalman Filter Predictions for Italy . . . . . . . . . . . . . . . . . 62
4.3.5 Kalman Filter Predictions for Singapore . . . . . . . . . . . . . 62
4.3.6 Model Performance Comparison Between Countries . . . . 66
4.4 Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5 Artificial Intelligence Prediction for the COVID-19 Data Based
on LSTM Neural Networks and H2O AutoML . . . . . . . . . . . . . . . . 69
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
5.2 Materials and Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.2.1 Long Short-Term Memory Networks (LSTM) . . . . . . . . . 70
5.2.2 H2O AutoML . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.2.3 Proposed Methodology for the Data Analysis . . . . . . . . . 71
5.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.3.1 LSTM Predictions for China. . . . . . . . . . . . . . . . . . . . . . 72
5.3.2 H2O AutoML Results for China. . . . . . . . . . . . . . . . . . . 72
5.3.3 LSTM Predictions for the United States . . . . . . . . . . . . . 76
Contents vii
5.3.4 H2O AutoML Results for US. . . . . . . . . . . . . . . . . . . . . 76
5.3.5 LSTM Predictions for Brazil. . . . . . . . . . . . . . . . . . . . . . 76
5.3.6 H2O AutoML Results for Brazil. . . . . . . . . . . . . . . . . . . 80
5.3.7 LSTM Predictions for Italy. . . . . . . . . . . . . . . . . . . . . . . 80
5.3.8 H2O AutoML Results for Italy. . . . . . . . . . . . . . . . . . . . 80
5.3.9 LSTM Predictions for Singapore. . . . . . . . . . . . . . . . . . . 80
5.3.10 H2O AutoML Results for Singapore. . . . . . . . . . . . . . . . 86
5.4 Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
6 Predicting the Geographic Spread of the COVID-19 Pandemic:
A Case Study from Brazil. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
6.2 Proposed Methodology for the Geographic Epidemic Predictor . . . 90
6.2.1 Location and Database . . . . . . . . . . . . . . . . . . . . . . . . . . 91
6.2.2 Clustering and Prediction Techniques . . . . . . . . . . . . . . . 92
6.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
6.3.1 General Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
6.3.2 Prediction Case 01—Analyzing the Infection Trend. . . . . 94
6.3.3 Prediction Case 02—Importance of the Prediction
After the Peak . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
6.4 Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
Chapter 1
Prediction for Decision Support During
the COVID-19 Pandemic
1.1 Introduction
At the moment this book is being written, the world is facing the most significant
challenge of modern history: the COVID-19 pandemic. The consequences of the
outbreakcannotbepreciselyevaluatedyet,sincedifferentcountriesaregoingthrough
differentstrategiestocopewiththeviruspandemicandthisis,obviously,generating
completelydifferentresults.Ontheotherhand,countrieswhereeffectivestrategies
couldreducetheinfectionratesafewmonthsago,arefacingsecondandthirdwaves
ofinfection,indicatingthatthiswillbealong-termbattlethatwillprobablyendnot
onlywhenavaccinewillbesuccessfullydeveloped,butwhentheimmunizationwill
beeffectivelyinplace,withbillionsofpeopleprotected.
Theimpactsarestronglysignificantinpublichealthsystems,withunprecedented
numbersofmortalityrates,exponentialincreasingratesofICUoccupationinshort
term and limited capacity of response in a large number of healthcare institutions.
Additionally, when considering the perspective of the consequences of the pan-
demicintheglobaleconomy,thefactsarealsofrightfulandwithlong-termimpacts,
especially for developing countries and more specifically for small and medium
enterprises.Societiesandgovernmentswillneedalongtimeandnewstrategiesand
actionstorecoverfromthecurrentoutbreakminimizingtheimpact.
ThetermCOVID-19isanacronymfor“CoronavirusDisease19”andisthename
ofthediseasecausedbythevirusstrainnamedSARS-CoV-2(SevereAcuteRespi-
ratory Syndrome Coronavirus 2), which belongs to the class of the coronaviruses.
Itattackstherespiratorysystem,butbesidesacommoninfectionintherespiratory
tract,severalotherconsequencesmayaffectthepatientdirectlyandindirectly.The
immunesystem,forexample,tryingtocombatthevirus,inseveralcasesgenerates
such a strong response that affects the patient’s life. Consequences of the cardio-
vascularsystemandheartfunctioningarealsobeingwidelyreportedinthemedical
literature.
WhentheviruswasfirstlyreportedinWuhan,Chinaandstarteditsexponential
spread,severalstrategieswereadoptedinparallel,ledbygeneralguidelinesprovided
by the WHO (World Health Organization) and specific efforts developed by local
©TheAuthor(s),underexclusivelicensetoSpringerNatureSwitzerlandAG2021 1
J.A.L.Marquesetal.,PredictiveModelsforDecisionSupport
intheCOVID-19Crisis,SpringerBriefsinAppliedSciencesandTechnology,
https://doi.org/10.1007/978-3-030-61913-8_1
2 1 PredictionforDecisionSupportDuringtheCOVID-19Pandemic
Governmentstodealandrespondtothischallenge,besidesthebiotechnologyefforts
todevelopmedicinesandvaccinestofightagainstthevirusaction[13].
From the biomedical engineering and computer science perspective, several
approaches are being considered, from the development of automatic diagnostic
toolsbasedontheapplicationofAI(ArtificialIntelligence)softwaretoanalyzeX-
ray and CT-Scans of the patient lungs, to the use of computer-based solution for
predictionusingepidemiologicaldatabasedonlinear,nonlinear,machinelearning,
anddeeplearningtechniques.
Themainobjectiveofthisbookistopresentdifferentapproachesandtechniques
for epidemiological time-series prediction, which will be able to provide decision
supportforGovernmentsandHealthcaredecision-makers,withthepresentationof
asequenceofresultsbasedonrealdatafromfivecountrieswithsomesimilarities
andsignificantdifferences.
The intent is not to exhaust the topic, but present a discussion covering from
conventionalcompartmentalmodels(SIRandSEIR)torecentlydevelopedArtificial
Intelligence (AI) solutions. The mathematical level presented for each technique
intends to be sufficient for the reader without an advanced technical background,
providingreferencesfortheoneswhowanttofollowupindetailabouteachapproach
presented.
Thisintroductorychapteraimstocreateafoundationofconceptsrelatedtothe
topicofpredictioninhealthcareandepidemiology,whicharerelevanttounderstand
theapplicationofthetechniquespresentedineachfollowingchapter.
1.2 ExplanationandPrediction
Atfirst,itisextremelyimportanttounderstandthedifferenceoftwoessentialcon-
cepts that are often mismatched: explanation, which will be clearly related to the
traditionalmethodspresentedinexplanatorystatistics;andprediction,whichisthe
applicationofspecifictechniquestopredictnewinformationfromcurrentdataand
willbethefocusofthisBook.
Thisdifferentiationisakeysteptounderstandwhenandwhythereisthenecessity
tousepredictiontechniquesandalsotoclearlyidentifywhatisthegoalandthetype
ofdatabeingconsidered.
Inferentialstatisticshavebeenwidelyusedinexplanatoryanalysisandthefocus
isusuallycausal[2].Theapplicationofdifferentregressiontechniquesarecommon,
accordingtotheresearchscopeandtypeofdata.Fourofthemostusedapproaches
arepresentedinSect.1.2.1.