Classification in high-dimensional data by extreme learning machine in R - r

Question

When I simulate 50 high-dimensional data set from multivariate normal distribution and classify by ELM, I find the AUC scores I get very different from each other in each loop. In some loops it goes below 50%. Where am I doing wrong? I want the deviation in the scores I get to be low for each loop. How can I achieve stable and high scores for each loop? How should I make changes in data generation? I look forward to your valuable contributions.
install.packages("MASS")
install.packages("stats")
install.packages("pROC")
install.packages("elmNNRcpp")
library(MASS)
library(stats)
library(pROC)
library(elmNNRcpp)
######################################################
# DATA SIMULATE FUNCTION
######################################################
# rm(list = ls())
generateData<- function(n,p) {
pr <- seq(0.80, 0.40, length.out = p)
pr[1] <- 1
covmat <- toeplitz(pr)
mu= rep(0,p)
X_ <- data.frame(mvrnorm(n, mu = mu, Sigma = covmat))
X <- unname(as.matrix(sample(X_)))
vCoef = rnorm(ncol(X))
vProb =exp(X%*%vCoef)/(1+exp(X%*%vCoef))
Y <- rbinom(nrow(X), 1, vProb)
mydata= data.frame(cbind(X,Y))
return(mydata)
}
######################################################
# SIMULATED DATA
######################################################
n <- 100
p <- 120
nsim <- 50
set.seed(123)
mydata <- list()
for (k in 1 : nsim ) {
data <- generateData(n , p)
# table(data[ncol(data)])
X <- data[-ncol(data)]
Y <- data[ncol(data)]
mydata[[k]] <- data
}
######################################################
# ELM CLASSIFICATION
######################################################
######################################################
# ELM CLASSIFICATION FUNCTION
######################################################
fELMCLASS <- function(x, col_names){
trainIndex <- sample(1:nrow(x), size=0.7*nrow(x))
trainSet <- x[trainIndex,]
testSet <- x[-trainIndex,]
xtrain <- as.matrix(trainSet[, 1:(length(trainSet)-1)])
ytrain <- as.matrix(trainSet[, length(trainSet)])
xtest <- as.matrix(testSet[, 1:(length(testSet)-1)])
ytest <- as.matrix(testSet[, length(testSet)])
model=elm_train(xtrain, ytrain, nhid=25 , actfun='relu')
pred.class=elm_predict(model,xtest,normalize=TRUE)
roc.model=roc(as.factor(ytest) ~ as.numeric(pred.class), direction=c("auto"))
performance_metrics <-t(data.frame("AUC" = roc.model$auc))
return(performance_metrics)
}
######################################################
######################################################
# SAVE RESULTS
######################################################
datalist = data.frame()
for (j in 1:nsim) {
data_ELM <- as.data.frame(mydata[[j]])
datalist <- rbind(datalist,fELMCLASS(data_ELM, "data"))
}
datalist