Synthetic data included in the scutr package for benchmarking.

syn_data.R

set.seed(01152021)
col.num <- 10 # number of non-class columns
cls.num <- 20 # number of classes
obs.num <- 10 # number of observations for first class
del.obs <- 10 # increase in observation per class
dis.sd  <- 5  # SD of class centers around 0

mtx <- matrix(nrow=0, ncol=col.num+1) # add a class col

for (cls in 1:cls.num){
    newmtx <- sapply(1:col.num,
                     function(x){
                         center <- rnorm(1, 0, sd=dis.sd)
                         rnorm(obs.num, mean=center)
                     })
    newmtx <- cbind(newmtx, rep(cls, nrow(newmtx)))
    mtx <- rbind(mtx, newmtx)
    obs.num <- obs.num + del.obs
}

imbalance <- as.data.frame(mtx)
names(imbalance)[col.num+1] <- "class"
imbalance$class <- as.factor(imbalance$class)

use_data(imbalance, overwrite=T)