1
$\begingroup$

I would like to ask if there is a metric in R and Python that serves not only for binary classification. I've found Matthews Correlation Coefficient works well in Python, but in R is only a binary version. Of course, I'm talking about "real" not just binary metrics - not something like "It's binary, but you can use one vs one or one vs all."

$\endgroup$
1
$\begingroup$

You mean this? MLR's performance measures

Performance measure suitable for the iris classification task listMeasures(iris.task)

1] "kappa" "multiclass.brier" "multiclass.aunp" [4] "multiclass.aunu" "qsr" "ber"
[7] "logloss" "wkappa" "timeboth"
[10] "timepredict" "acc" "lsr"
[13] "featperc" "multiclass.au1p" "multiclass.au1u" [16] "ssr" "timetrain" "mmce"

Caret does too but fewer.

$\endgroup$
2
  • $\begingroup$ I asked for the multi-class classification and not the multi label classification. $\endgroup$ – atos Nov 28 '17 at 15:48
  • $\begingroup$ Please do not change your post when someone has already written an answer. Going to the bottom, the cohen kappa score is what I needed. Thanks ! $\endgroup$ – atos Nov 28 '17 at 21:31
1
$\begingroup$

How about the MLmetrics package in R? It has multi class log loss function that you can use to assess the performance of your model.

$\endgroup$
1
  • $\begingroup$ This solution is good. But I'm looking for something where inputs would be just categories not probability matrices. $\endgroup$ – atos Nov 28 '17 at 15:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.