auc_tidiers {broom} | R Documentation |
Tidy "roc" objects from the "auc" package. This can be used to, for example, draw ROC curves in ggplot2.
## S3 method for class 'roc' tidy(x, ...)
x |
an "roc" object |
... |
Additional arguments, not used |
A data frame with three columns:
cutoff |
The cutoff of the prediction scores used for classification |
tpr |
The resulting true positive rate at that cutoff |
fpr |
The resulting false positive rate at that cutoff |
If the labels had names, those are added as an "instance" column.
if (require("AUC", quietly = TRUE)) { data(churn) r <- roc(churn$predictions,churn$labels) td <- tidy(r) head(td) library(ggplot2) ggplot(td, aes(fpr, tpr)) + geom_line() # compare the ROC curves for two prediction algorithms library(dplyr) library(tidyr) rocs <- churn %>% tidyr::gather(algorithm, value, -labels) %>% group_by(algorithm) %>% do(tidy(roc(.$value, .$labels))) ggplot(rocs, aes(fpr, tpr, color = algorithm)) + geom_line() }