Home
ROC and AUC, Clearly Explained!
Sem categoria

ROC and AUC, Clearly Explained!

36 Comentários




* O link do vídeo está no final desse post.

Descubra tudo sobre Robô IQ OPTION de AUTOMATIZAÇÃO de Lista de SINAIS
E conheça também sobre o Melhor Bot Para Opções Binárias 2022
Que tal um robo gratuito para IQ Option? Sistema Para IQ Option Grátis





ROC (Receiver Operator Characteristic) graphs and AUC (the area under the curve), are useful for consolidating the information from a ton of confusion matrices into a single, easy to interpret graph. This video walks you through how to create and interpret ROC graphs step-by-step. We then show how the AUC can be used to compare classification methods and, lastly, we talk about what to do when your data isn’t as warm and fuzzy as it should be.

NOTE: This is the 2019.07.11 revision of a video published earlier.

NOTE: This video assumes you already know about
Confusion Matrices…

…Sensitivity and Specificity…

…and the example I work through is based on Logistic Regression, so it would help to understand the basics of that as well:

For a complete index of all the StatQuest videos, check out:

If you’d like to support StatQuest, please consider…
Patreon:
…or…
YouTube Membership:

…a cool StatQuest t-shirt or sweatshirt:

…buying one or two of my songs (or go large and get a whole album!)

…or just donating to StatQuest!

Lastly, if you want to keep up with me as I research and create new StatQuests, follow me on twitter:

0:00 Awesome song and introduction
0:48 Classifying samples with logistic regression
4:03 Creating a confusion matrices for different thresholds
7:12 ROC is an alternative to tons of confusion matrices
13:44 AUC to compare different models
14:28 False Positive Rate vs Precision
15:38 Summary of concepts

#statquest #ROC #AUC

Vídeo

Gosta de dicas de futebol para investimento desportivo? Então visita agora mesmo Palpites Copa do Mundo e prepara-se para viver a emoção do maior campeonato do planeta.





Tags:, , , , , , , , ,

You may also like

About The Author

estilominimalista

36 Comentários

  1. NOTE: This revised video corrects three minor errors (that, fortunately, did not affect the main idea) found by vigilant viewers in the original. Other than that, it's the same as before. That said, there is a minor error at 12:00 . The confusion matrix should be TP = 3, FP = 2, FN = 1, TN = 2. The displayed matrix should be for the next point.

  2. I've just upvoted the intro song. That's it. Oh wait, the video also explains the ROC and AUC? that's great!

  3. "AUC" is probably the most undescriptive acronym I have never heard of. Who comes up with these names? Well clearly they were gifted more mathematically than lingually since the idea is pretty cool.

  5. Love the content, but having ads every 2 minutes is a bit too much. If you could push them towards the end would be much better

  7. Honestly I had zero knowledge of any theory or terminology or even what this subject is and I’m baked as fuck and I now have a basic understanding of this concept. A fine job sir

  12. Thank you so much for this amazing explanation. This is the first time I actually understood the idea behind AUC and ROC.

  15. Don't know if I'm the only one, I love the content, it's very high-quality and well-explained. Just the opening guitar music makes me uncomfortable. I would appreciate shorter videos with clear-cut content introduction.

  20. It's confusing…no pun intended…especially when dealing with health care or anything else where a score of 1, is actually the negative outcome or the outcome you do not want to occur. So in this dataset, "is obese" is represented by 0 and "is not obese" is 1?

  22. Thank you for the wonderful work Josh! Just bought a StatQuest Triple BAM shirt. Is the threshold arbitrarily set to change by increments ? If so, what is the typical increments for thresholds?

  24. Thanks for the explanation.
    I am planning to go throw all the videos on the machin leaning playlist as it seems to have a lot of valuable informations.
    I can take it in 3 or 4 days but I want to implement and test these concepts using real world machine learning dataset and Scikit-Learn
    This will make it takes about 8 to 10 days! Excited to achieve this goal 😊

  25. Maybe I have a stupid quastion but… I'm wondering how do you colorize the area under the curve for ROC in R? I've managed to take only the value for auc (with [email protected][[1]]). And very useful video by the way 🙂

  26. This is an amazing explamation. Thank you for that. I have a question though. When we talk about object detection, which is a classification problem, hwo do we deal with true negatives? It could be a very large number. So do we use the false positive rate for object detection like YOLO or do we use something else?

  34. Thank you, Josh. Your videos are great. Do you by any chance have something on Cohen's Kappa and TSS? Cheers.