pntd.0010071.g002.tif (475.13 kB)
A comparison of Random Forest- and Lasso-based model performance.
figure
posted on 2022-01-25, 08:55 authored by Gal Koplewitz, Fred Lu, Leonardo Clemente, Caroline Buckee, Mauricio SantillanaThe mean is taken across the different cities, with the fill range of delays in availability of epidemiological information (from eight weeks, AR8, to one week, AR1) and the different feature sets (AR, GT AR+GT, AR+GT+W) shown.
History
Usage metrics
Categories
Keywords
finer spatial resolutionsassume short delaysassume long delaysaggregated weather variables6 &# 82111 &# 8211weekly incidence countsframework achieves meaningful3 weeks ),multiple data sourcesdifferent data sourcesproduce accurate predictionscombat dengue infectionsaedes aegypti mosquitobased data sources20 different citiesdespite significant effortsbrazil data sources20 citiesdifferent demographicepidemiological datamethodological frameworkdengue incidence8 weeksxlink ">use towardstime predictionsstrongest predictorsseasonal autocorrelationreliable vaccinerelative contributionrandom forestprior yearsprediction systemsmuch remainslittle worklevel predictionlasso regressionimproved predictionflourishing acrosserror rateeffective treatmentdominant predictorsdifficulties inherentcase studybased models
Licence
Exports
RefWorks
BibTeX
Ref. manager
Endnote
DataCite
NLM
DC