1 / 3100%
Theh integrativeh approachh toh datah analysish providesh insighth abouth theh
selectedh predictiveh modelsh forh development.h h Thish developmenth
createsh ah connectionh toh presenth ah sectionh associatedh withh theh modelh
descriptionh forh interpretationh ofh eachh ensembleh nodeh withh
combinations.h h Theh ensembleh nodeh withh combinationh includesh theh
investigationh ofh h average,h maximum,h andh votingh methodsh inh SASh
Enterpriseh Miner.h h Theh combiningh methodsh usingh ah decisionh tree,h
logistich regression,h andh neuralh networkh modelsh canh beh introduceh forh
predictiveh modelsh withh comparison.h h Theseh modelingh resultsh createsh
ah comparisonh ofh heterogeneoush practicesh toh trainh anh intervalh targeth
andh classh targeth forh evaluation.h h Theseh methodsh areh designedh toh
determineh theh intervalh targetsh andh toh describeh theh predictedh valuesh
ofh averagesh andh maximumh conditionsh withh theh ensembleh nodes.
Theh datah analysish andh theh relationshipsh betweenh datah toolsh canh
addressh featuresh andh theh targeth variablesh ash thish sectionh willh provideh
inh detail.h h Theh targeth variableh withh visualizationh andh predictiveh
modelsh ofh Rh Language,h SASh Enterpriseh Miner,h andh h Weka,h andh
AutoMLh softwareh canh influenceh theh predictionh ofh types.h h h Theh
outcomeh ofh theh overallh findingsh accordingh toh targeth andh typesh areh
featuresh thath willh provideh additionalh insighth relatedh toh theh evidenceh
andh observationh ofh theh trainingh results.h h Theh implementedh predictiveh
modelingh approachh usingh theseh toolsh includesh buth noth limitedh toh theh
fith model,h targeth correlation,h ROCh curve,h correlationh matrixh andh
otherh modelingh outputs.h h Theseh modelingh outputsh areh usedh forh
evaluationh toh assessh theh parametersh designedh ash featuresh forh
comparisonh andh datah investigation.h h Theh comparisonh ofh theh datah seth
withh anh introductionh ofh varioush datah analysish toolsh canh aidh withh theh
datah findingsh byh obtainingh driversh toh determineh desiredh factorsh forh
anh investigativeh process.h h Theh resultsh canh alsoh rankh withh comparisonh
ofh theh visualizationh andh predictiveh modelsh ofh theh toolsh identifiedh forh
validationh purposes.
Theh datah analysish toolsh thath areh beingh consideredh willh allowh forh theh
evaluationh detailsh toh developh ah predictiveh analyticsh design.h h Theseh
detailsh toh developh ah predictiveh modelh approachh withh theh
combinationh ofh theh ensembleh nodesh canh beh appliedh forh
interpretationh usingh theh investigationh ofh interactionh betweenh flighth
maneuvers.h h Theh datah analysish toolsh willh beh appliedh toh assessh theh
abilityh toh performh flighth maneuversh andh piloth datah ofh steep,h normal,h
andh swallowh performancesh ofh rotorcrafth systems.h h Theseh
environmentsh willh beh addressingh theh transitionalh phasesh ofh rotorcrafth
systemsh andh theh flighth datah withh performanceh ofh theh pilot’sh abilityh toh
executeh theh maneuver.h Theh datah analysish toolsh areh mentionedh
accordingh toh theh investigativeh frameworkh usingh Rh Language,h SASh
Enterpriseh Miner,h Weka,h andh othersh suchh ash AutoMLh (seeh figureh
belowh onh theh environmenth interaction).
Figure
.h h Datah analysish designedh forh DATAh 670h Projecth withh theh
Federalh Aviationh Administrationh (FAA)h ofh toolsh forh evaluationh
purposes.
Theh datah analysish visualizationh offersh howh theh datah analysish toolh canh
serveh theh interpretationh ofh theh studyh findings.h h However,h theh needh
toh identifyh ah classh targeth inh thish studyh willh beh vitalh toh theh datah
solutionh andh toh theh investigativeh framework.h h Theh FAAh datah inh
creatingh anh integrativeh approachh alsoh presentsh theh needh toh addressh
theh performanceh maneuversh accordingh toh flighth characterizationh forh
classificationh (e.g.,h steep,h normal,h andh swallow).h h Fromh theh classh
target,h theh associationh ofh eachh predictedh valueh canh beh usedh toh revealh
theh methodh ofh performanceh usingh ensembleh models.h h Forh example,h
theh useh ofh classh targetsh findingsh withh theh methodsh ofh average,h
maximum,h andh votingh (withh theh votingh posteriorh probabilitiesh resultsh
toh theh averageh forh eachh approach).h h Inh consideration,h thish approachh
highlightsh theh abilityh toh deployh theh necessaryh datah analysish
visualizationh andh predictiveh modelsh toh improveh performanceh andh toh
buildh onh theh piloth experienceh forh anh integrativeh approachh thath ish
betterh suitedh forh complexh environmentsh withh theh useh ofh datasets.
Students also viewed