It implements the ensemble quantile classifier (EQC) that extends the quantile-based classifier (Hennig and Viroli, 2016b). An accompanying vignette illustrates a simulation study and the application of the EQC compared to the other methods.
To install this package, run
if (!require("devtools")) install.packages("devtools")
devtools::install_github("CliffordLai/eqc")
# To include the vignette, one can use the following
# but it takes much more time to install the package
devtools::install_github("CliffordLai/eqc",build_vignettes = TRUE)This is a basic example which shows you how to train and tune the EQC.
library(eqc)
# Divide data into training set and test set randomly
data(wdbc)
set.seed(193)
trainIndex <- sample(c(rep(TRUE,2),rep(FALSE,1)),nrow(wdbc),replace=TRUE)
train <- as.matrix(wdbc[trainIndex,-1])
cl.train <- wdbc[trainIndex,1]
test <- as.matrix(wdbc[!trainIndex,-1])
cl.test <- wdbc[!trainIndex,1]
p <- ncol(train)
# Tuning parameters
thetaList <- rbind(rep(0.3,p),rep(0.4,p),rep(0.5,p))
# Fit the tuned EQC with a ridge logistic regression
lambda <- c(3,1,0.1,0.01,0.001,0.0005,0.0001)
fit <- eqcTrain(train,cl.train,
thetaList=thetaList,
method = "glmnet",
alpha = 0,lambda = lambda,
tuneControl = list(nfolds=5,fold.seed=117),
lower.limits=0, upper.limits=Inf)
fit$cvparameter #Selected tuning parameters
#> $thetaCV
#> [,1] [,2] [,3] [,4] [,5] [,6] [,7] [,8] [,9] [,10] [,11] [,12] [,13]
#> [1,] 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5
#> [,14] [,15] [,16] [,17] [,18] [,19] [,20] [,21] [,22] [,23] [,24]
#> [1,] 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5
#> [,25] [,26] [,27] [,28] [,29] [,30]
#> [1,] 0.5 0.5 0.5 0.5 0.5 0.5
#>
#> $lambdaCV
#> [1] 0.01
#>
#> $costCV
#> [1] NA
acc <- mean(predict(fit,newdata = test,type = "class")[[1]]==cl.test)
acc #Classification accuracy
#> [1] 0.9772727The vignette include a simulation study and an application. List vignettes in an HTML browser by,
browseVignettes(package = "eqc")