plot_clf_feature_selection: plot_clf_feature_selection

In shihuang047/crossRanger: Cross-Application of Ranger Models using microbiome data

plot_clf_feature_selection

Description

Plot the classification performance against the gradually reduced number of features used in the modeling.

Usage

plot_clf_feature_selection(
  x,
  y,
  nfolds = 5,
  rf_clf_model,
  metric = "AUROC",
  positive_class = NA,
  outdir = NULL
)

Arguments

x	The data frame or data matrix for model training.
y	A factor related to the responsive vector for training data.
nfolds	The number of folds in the cross-validation for each feature set.
rf_clf_model	The rf classification model from rf.out.of.bag
metric	The classification performance metric applied. If binary classification, this must be one of "AUROC", "Accuracy", "Kappa", "F1". If multi-class classification, this must be one of "Accuracy", "Kappa".
positive_class	A class of the y.
outdir	The output directory.

Author(s)

Shi Huang

Examples

set.seed(123)
require("gtools")
n_features <- 100
prob_vec <- rdirichlet(5, sample(n_features))
x <- data.frame(rbind(t(rmultinom(7, 7*n_features, prob_vec[1, ])),
            t(rmultinom(8, 8*n_features, prob_vec[2, ])),
            t(rmultinom(15, 15*n_features, prob_vec[3, ])),
            t(rmultinom(15, 15*n_features, prob_vec[4, ])),
            t(rmultinom(15, 15*n_features, prob_vec[5, ]))))
y<-factor(c(rep("A", 30), rep("C", 30)))
s<-factor(rep(c("B1", "B2", "B3", "B4"), 15))
rf_model<-rf.cross.validation(x, y, nfolds=5)
summ <- plot_clf_feature_selection(x, y, nfolds=5, rf_model, metric="AUROC", outdir=NULL)
summ <- plot_clf_feature_selection(x, y, nfolds=s, rf_model, metric="AUROC", outdir=NULL)
res<-replicate(10, plot_clf_feature_selection(x, y,
nfolds=5, rf_model, metric="AUROC", outdir=NULL))
do.call(rbind, res["top_n_perf", ])

shihuang047/crossRanger documentation built on Nov. 8, 2024, 2:49 a.m.