bestNN: Function to select the best nnet, but this could be in the...

Description Usage Arguments

Description

nnet is obtained with all vars, so-called oob error 1 - cor(predict(Nnet)[complete.cases(predict(Nnet)),1], Y[complete.cases(cbind(Y, xdata[, Nnet$coefnames]))])^2) an iterated process consists on creating a new nnet with the (1-vars.drop.frac)*vars more important obtained in the previous step and generating a new nnet model. The process stops when abs(mean(oob error)+sd(oob error)) is >= than previous step or only two vars are left

Usage

1
bestNN(xdata, Y, size = 2, vars.drop.frac = 0.2, verbose = FALSE)

Arguments

xdata

data set to be analyzed

Y

var to analyze

size

nnet param: number of units in the hidden layer. Can be zero if there are skip-layer units

vars.drop.frac

fraction of variables removed at each iteration step, it selects the most important variables 1-fraction at each step


isglobal-brge/nlOmicAssoc documentation built on May 9, 2019, 8:16 a.m.