基于多阈值前馈神经网络的症状和气候疟疾威胁检测Symptomatic-and-Climatic...

I.J.InformationTechnologyandComputerScience,2017,8,40-47PublishedOnlineAugust2017inMECS()DOI:10.5815/ijitcs.2017.08.05Copyright©2017MECSI.J.InformationTechnologyandComputerScience,2017,8,40-47SymptomaticandClimaticBasedMalariaThreatDetectionUsingMultilevelThresholdingFeed-ForwardNeuralNetworkAbisoyeOpeyemiA.1FederalUniversityofTechnology,DepartmentofComputerScience,SchoolofinformationandCommunicationTechnology(SICT),P.M.B.65,Minna,NigerState,Nigeria.E-mail:[e-mail],[e-mail].2UniversityofIlorin,Ilorin,DepartmentofComputerScience,FacultyofCommunicationandInformationScience(FCIS),P.M.B.1515,Nigeria.E-mail:[e-mail],[e-mail]:21February2017;Accepted:09May2017;Published:08August2017Abstract—Recentworldwidemedicalresearchisfocusingonnewintelligenceapproachesfordiagnosisofvariousinfections.Thesporadicoccurrenceofmalariadiseasesinhumanhaspushedtheneedtodevelopcomputationalapproachesforitsdiagnoses.Mostexistingconventionalmalariamodelsforclassificationproblemsexaminethedynamicsofasymptomaticandmorphologicalcharacteristicsofmalariaparasiteinthethickbloodsmear,butthisstudyexaminethesymptomaticcharacteristicsofmalariaparasitecombinedwitheffectsofclimaticfactorwhichisagreatdeterminantofmalariaseverity.Theneedtopredicttheoccurrenceofmalariadiseaseanditsoutbreakwillbehelpfultotakeappropriateactionsbyindividuals,WorldHealthOrganizationsandGovernmentAgenciesanditsdevastatingimpactwillbereduced.ThispaperproposedFeed-ForwardBack-Propagation(FF_BP)NeuralNetworkmodeltodeterminetherateofmalariatransmission.Monthlyaveragesofclimaticfactors;rainfall,temperatureandrelativehumiditywithmonthlymalariaincidenceswereusedasinputvariables.Anoptimumthresholdvalueof0.7100withclassificationaccuracy87.56%,sensitivity96.67%andspecificity76.67%andmeansquareerrorof0.100wereachieved.While,themodelmalariathreatdetectionratewas87.56%,positivepredictivevaluewas89.23%,negativepredictivevaluewas92.00%andthestandarddeviationis2.533.ThestatisticalanalysisofFeed-ForwardBack-PropagationNeuralNetworkmodelwasconductedanditsresultswerecomparedwithotherexistingmodelstocheckitsrobustnessandviability.IndexTerms—Malaria,Feed-ForwardBack-PropagationNeuralNetwork(FF_BP),Classification,Symptomatic,Climatic,Multiclass,MultilevelThresholding.I.INTRODUCTIONSeveraldiseasesaffectstheproperfunctioningofhuman‗shealthandpoorhealthconditionaffectsone‘slifespanandachievements.Oneofsuchcommondiseaseismalaria.AccordingtoWorldHealthOrganization(WHO)Reportin2015,thereare95countriesandterritorieswithongoingmalariatransmission,andonly6countrieshaveeliminatedmalaria(WHO,2015).Globally,anestimated3.4billionpeopleareatriskofmalaria.WHOestimatesthat214millionnewcasesofmalariaoccurredgloballyin2015(range149-303million)andanestimateddeathof438,000,000.Mostcases(88%)anddeaths(90%)occurredinAfrica,andmostdeaths(71%)wereinchildrenunder5yearsofage(WHO,2014).WHOhasmademalariareductionachiefpriority.Theideaistodisablethediseasebycombiningvirtuallyeveryknownmalaria-fightingtechnique.BillGates,whohascalledmalaria―theworstthingontheplanet,‖hasdonatedhundredsofmillionsofdollarstotheeffortthroughtheBillandMelindaGatesFoundation.TheBushAdministrationhaspledged1.2billiondollars(WHO,2015).Fundsdevotedtomalariahavedoubledsince2003.Theideaistodisablethediseasebycombiningvirtuallyeveryknownmalaria-fightingtechnique,fromtheancient(Chineseherbalmedicines)totheold(bednets)totheultramodern(multidrugcocktails).But,therateatwhichmalariaprevailsintheconcernedareasseemsexponentialdespitethefactthatallhandsareondesktocombatthespreadandeveneradicatethedisease(OwoseniandOgundahunsi,2016).Atthesametime,malariaresearchersarepursuingalong-sought,elusivegoal:avaccineandmethodsthatwouldcurbthediseaseforgood(WHO,2014;2015).SymptomaticandClimaticBasedMalariaThreatDetectionUsingMultilevel41ThresholdingFeed-ForwardNeuralNetworkCopyright©2017MECSI.J.InformationTechnologyandComputerScience,2017,8,40-47Malariatransmissionissitespecificduetodifferentclimaticchangeofaregion.Therearethreemainclimaticfactorsthatdirectlyaffectmalariatransmission.Theyaretemperature,rainfallandrelativehumidity(Depinayetal,2004).Severalnon-climaticfactors,suchashumanorbehaviouralfactorscanalsoaffectthepatternofmalariatransmissionandtheseverityoftheproblem.Malariaparasitedevelopsmorequicklyathighertemperatures.Highertemperaturesincreasethenumberofbloodmealstakenandthenumberofeggslaidbythemosquitoes,whichincreasesthenumberofmosquitoesinagivenarea.Theminimumtemperatureformosquitodevelopmentisbetween8–10°C;theoptimumtemperatureis25–27°C,andthemaximumtemperatureis40°C.Altitudeinfluencesthedistributionandtransmissionofmalariaindirectly,throughitseffectontemperature.Asaltitudeincreases,temperaturedecreases,sohighlandsarecolderandlowlandsarewarmer.Datamininghasgreatpotentialforthehealthcaresystemstousemedicaldataforanalysisandtoofferimprovedhealthcareatreducedcost.Theseinhealthcareplayasignificantroleinpredictionanddiagnosisofvarioushealthproblemslikehear