###### Install RNeat
# install.packages("devtools")
# library("devtools")
# install_github("RNeat","ahunteruk")
library("RNeat")


trainSamples <- sample(nrow(iris), round(nrow(iris)*0.5))
validationSamples <- setdiff(1:nrow(iris), trainSamples)

data(iris)
iris[,5] <- unclass(factor(iris[,5]))      #convert classes to numbers
targetClasses <- max(iris[,5])
iris[,5] <- (iris[,5]-1)/(targetClasses-1) #rescale target to [0;1]

rneatsim <- rneatneuralnet(formula = Species ~ Sepal.Length + Sepal.Width + Petal.Length + Petal.Width,
                           trainingData = iris[trainSamples, ],
                           nTrainingGenerations = 20
						   )

results <- compute(rneatsim,iris[validationSamples, 1:4])
results[,5] <- round(results[,5] * (targetClasses-1))/(targetClasses-1) #round result to required levels

results[,5] == iris[validationSamples, 5]


cat("Validation results:", 100*round(mean(results[,5] == iris[validationSamples, 5]),2),"% are correct", sep="")

plot(rneatsim)
dev.new()
plot(rneatsim$simulation)