Dear R-philes,
I am having some trouble averaging across multiple runs of a
classifier in an ROC Curve. I am using the ROCR package and the
plot() method.
First, I initialize a list with two elements where each element is a
list of predictions and labels:
vowel.ROC <- list(predictions=list(), labels=list())
For every run of the classifier, I append the scores and labels to
their corresponding list elements as follows:
vowel.ROC$predictions <- append(vowel.ROC$predictions,
list(scores))
vowel.ROC$labels <- append(vowel.ROC$labels, list(numGrps))
The R display of vowel.ROC looks like the following (not sure why NULL
is there):
List of 2
$ predictions:List of 4
..$ : num [1:148] 0.234 0.293 0.275 0.391 0.191 ...
..$ : num [1:152] 0.99 0.974 0.934 0.767 0.934 ...
..$ : num [1:293] 0.05009 0.00739 0.03211 0.85894 0.18265 ...
..$ : num [1:247] 0.0184 0.0168 0.0942 0.0149 0.089 ...
$ labels :List of 4
..$ : num [1:148] 1 1 1 1 1 2 2 1 1 1 ...
..$ : num [1:152] 2 2 2 2 2 1 1 1 1 2 ...
..$ : num [1:293] 1 1 1 1 2 1 1 2 2 1 ...
..$ : num [1:247] 1 1 1 1 1 2 1 1 1 2 ...
NULL
I make prediction and performance objects and plot the resulting
performance object as shown below.
pred <- prediction(vowel.ROC$predictions, vowel.ROC$labels)
perf <- performance(pred, "tpr", "fpr")
plot(perf, avg="horizontal", spread.estimate="boxplot")
Instead of getting an average ROC curve, I am getting four separate
ones on the same plot. This is nice but I prefer the average curve.
Any direction would be much appreciated.
Regards,
Na'im
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
