roc_plot: Plot the Receiver Operating Characteristic curve for preds...

In jpgard/abroca: Compute and Evaluate Absolute Between-ROC Area

Description

Plot the Receiver Operating Characteristic curve for preds and labs.

Usage

roc_plot(preds, labs, plot_type = NULL, show_diag = FALSE,
  fout = NULL)

Arguments

preds	vector of predictions
labs	vector of true class labels
plot_type	if set to "majority" or "minority", coloring is set to match plots generated by slice_plot (blue for majority class, red for minority); if NULL then neutral coloring is used
show_diag	boolean indicator for whether a diagonal line indicating the random chance ROC curve should be plotted
fout	path to an output file (png)

Value

No return value; file is saved to disk.

Examples

 
# First, we load data, train a model, and generate predictions to evaluate.
data("recidivism")
recidivism$returned = as.factor(recidivism$Return.Status != "Not Returned")
in_train = caret::createDataPartition(recidivism$returned, 
    p = 0.75, list = FALSE)
traindata = recidivism[in_train,c("Release.Year", "County.of.Indictment", 
    "Gender", "Age.at.Release", "returned")]
testdata = recidivism[-in_train,c("Release.Year", "County.of.Indictment", 
    "Gender", "Age.at.Release", "returned")]
lr = glm(returned ~ ., data=traindata, family="binomial")
testdata$pred = predict(lr, testdata, type = "response")

# Now, we apply roc_plot() to the labels and predictions 
# (note that this writes a file to fout):
roc_plot(testdata$pred, testdata$returned, plot_type = "majority", 
    show_diag=TRUE, fout="roc.png")

jpgard/abroca documentation built on May 25, 2019, 11:31 p.m.