Use AUC to evaluate multiclass problems

AUC is an excellent evaluation metric for binary classification, especially if you have class imbalance.
New in scikit-learn 0.22: AUC can be used with multiclass problems! Supports "one-vs-one" and "one-vs-rest" strategies.
👉 New tips every TUESDAY and THURSDAY! 👈
🎥 Watch all tips: • scikit-learn tips
🗒️ Code for all tips: github.com/jus...
💌 Get tips via email: scikit-learn.tips
=== WANT TO GET BETTER AT MACHINE LEARNING? ===
1) LEARN THE FUNDAMENTALS in my intro course (free!): courses.datasc...
2) BUILD YOUR ML CONFIDENCE in my intermediate course: courses.datasc...
3) LET'S CONNECT!
- Newsletter: www.dataschool...
- Twitter: / justmarkham
- Facebook: / datascienceschool
- LinkedIn: / justmarkham

[@dataschool]

New to ROC curves and AUC? Check out my 14-minute explanation: ruclips.net/video/OAl6eAyP-yo/видео.html (it's my 2nd most popular video!)

[@cgmiguel]

It would be nice to hear about your thoughts on AUC ROC vs AUC PR, for imbalanced sets (or any other metic for that matter). Thanks for sharing!

[@dataschool]

Great question! I've got a long lesson about that topic in my upcoming ML course, but unfortunately I can't boil it down to a few sentences. Stay tuned!

[@felixisnr1]

thank you for all your vids, really appreciated.
do you plan to do a short series for plotting? would love to see this on your channel if you find the time to do so at one point.
either way, thanks a lot. you are such a great teacher!

[@dataschool]

Thanks for your suggestion, and for your kind words! I don't have a series on plotting planned, but I'll certainly consider it for the future!

[@priyabrataray8663]

Really Helpful 🙂.Can you please suggest a demo dataset for multi class classification

[@dataschool]

Glad it's helpful! You could try the iris, digits, or wine datasets: scikit-learn.org/stable/datasets/toy_dataset.html

[@debabratasikder9448]

axis 1 is out of bounds for array of dimension 1

[@beautyisinmind2163]

So how to decide which one should i use actually ovo or ovr if i have 4 classes to be classified

[@svitirur1665]

nice, good explanation .
I have question can we do binomial /poisson test experiment on a dataset?

[@dataschool]

You should be able to do it using SciPy. Hope that helps!