R- Deep Learning.R

closeAllConnections()
rm(list=ls())
setwd("/Volumes/16 DOS/R_nbs")
logitML<-read.csv("LogitML2.csv",sep=",",header=TRUE,fileEncoding="latin1")
logitML<-logitML[1:192,]
library(ISLR)
library(neuralnet)
library(caTools)

set.seed(1500)

padron <- as.data.frame(scale(logitML[,c(18:29,31)],center = apply(logitML[,c(18:29,31)], 2, min), scale = apply(logitML[,c(18:29,31)], 2, max) - apply(logitML[,c(18:29,31)], 2, min)))

c0_recomend = as.numeric(logitML$c0_recomend)
data = cbind(c0_recomend,padron)

split = sample.split(data$c0_recomend, SplitRatio = 0.80)
train = subset(data, split == TRUE)
test = subset(data, split == FALSE)
vars <- names(padron)

for (i in c(1,7)){
  print(which(names(test)==vars[i]))}

gg <- paste(feats,collapse=' + ')
gg <- paste('c0_recomend ~',gg)
gg <- as.formula(gg)
neural <- neuralnet(gg,data,hidden=c(4,4),linear.output=TRUE,learningrate = 0.01, stepmax=10000)

predicted<- compute(neural,data[,2:14])
predicted$net.result-data[,1]
1-abs(mean(abs(predicted$net.result)-data[,1]))

weights<-predicted.nn.values$neurons

par(mfrow=c(1,1))
par(mar=c(4, 2, 2, 2))

### Requires Quartz to plot
plot(nn)