Hands down the best intuitive explanation of the concept of ROC. Not even the gold standard books has taken the pain to explain this way. You inspire us!
Thank you so much! You are a MASTER Teacher. It is very hard to find teachers who teach in a way that makes sense and you have that gift. I LOVE your channel and wish you tons of success. I am an MIT student and will share it with my friends.
I'd be a sinner if I would just walk away after watching such an amazing video! I'd want this video to be shown first on YT when someone searches for 'ROC curve'. Thanks a lot for this, mate!
Your format of hand-drawn presentation is amazing, it helps tremendously with understanding complex topics! Keep up the great work, it is much appreciated. :)
Love it! The only thing I'd add is clarify the underlying basic assumption that the predictions of the model are spread across this 0-1 probability space with a lot of mass on either ends and decreasing mass as you go to the center but more mixing.
Your channel is the best, thank you. I got stuck in my learning, but then found answers here (a few times already). Think I will just come here directly next times. 😄
'This is the best ROC explainer video that I have found to date...And, I have watched a lot! Thanks so much. Very clear presentation of key concepts illustrated with a realistic example. The pre-prepared hand drawn whiteboard illustrations were brilliant and easy to understand as your went through the video. I think I might use that technique in my own teaching vs drawing as I go. So much more efficient.
this was a very well explained video. thanks so much. this concept is not usually taught that well to be honest so you cleared up a lot of confusion that I Had.
Good video, I also think of ROC Curve as a visual summary tool to help select the most suitable threshold, without having to go through lots of different confusion matrices.
Good question. We can still use this ROC curve idea and extend to multiclass. Say we have three classes {0,1,2} and suppose class 1 and class 2 we care about most. Then, we can build one ROC curve for class 0 vs class 1 and one for class 0 vs class 2 and try to optimize both of these.
Hi Ritvik. Thank you for the great content and really loved your explanation on ROC curve-v helpful. Quick question-so why do we use Precision Recall instead of ROC curve. Do you have video on the same?. I read Precision Recall curve is used more in imbalanced dataset scenario-any inputs here? Thank you:)
Why do we use the area under the curve vs using the minimum distance from the star point (top left corner) to the ROC? Is there any official reason for that?
The connection to the business scenario really helps to understand the purpose and movements of the ROC... Thank you for your excellent explanation.
Glad it was helpful!
Hands down the best intuitive explanation of the concept of ROC. Not even the gold standard books has taken the pain to explain this way. You inspire us!
Thank you so much! You are a MASTER Teacher. It is very hard to find teachers who teach in a way that makes sense and you have that gift. I LOVE your channel and wish you tons of success. I am an MIT student and will share it with my friends.
Wow, thank you!
I'd be a sinner if I would just walk away after watching such an amazing video! I'd want this video to be shown first on YT when someone searches for 'ROC curve'. Thanks a lot for this, mate!
Watched lot of other videos, I got ROC concept 100% without any doubts from this video. Easy to understand. appreciate you effort.
Great to hear!
This is hands down THE BEST explanation of RoC construction, dynamic and interpretation I have ever viewed.
This is gold, you deserve way more subscribers! Thank you for using a business case, it makes it easier to understand.
Your format of hand-drawn presentation is amazing, it helps tremendously with understanding complex topics! Keep up the great work, it is much appreciated. :)
Thank you very much!
You are literally the best teacher for data science on UA-cam!
Glad you think so!
Love it! The only thing I'd add is clarify the underlying basic assumption that the predictions of the model are spread across this 0-1 probability space with a lot of mass on either ends and decreasing mass as you go to the center but more mixing.
Your channel is outright amazing!! Discovered your channel recently, you explain things in such a short, simple and intuitive manner
Awesome, thank you!
i can watch that the whole night!
this is awesome!
Best Explanation of the ROC curve I have seen on UA-cam so far!
Your channel is the best, thank you.
I got stuck in my learning, but then found answers here (a few times already). Think I will just come here directly next times. 😄
Happy to help!
Very well explained. The drawings are really helping to uncover the intuition behind them.
Excellent teaching! I love to understand how things work and you does a good job on it.
one of the best channels ever. very informative straight to the point and takes my understanding to a whole new level. I cant thank you enough
'This is the best ROC explainer video that I have found to date...And, I have watched a lot! Thanks so much. Very clear presentation of key concepts illustrated with a realistic example. The pre-prepared hand drawn whiteboard illustrations were brilliant and easy to understand as your went through the video. I think I might use that technique in my own teaching vs drawing as I go. So much more efficient.
Your explanations are absolutely the best.
One of the best channels with very intuitive and structured explanations. I wish I knew about you during my university study 👨🎓
Thank you!
you are truly a gifted teacher, thank you so much for helping me getting ready for my interviews!
I agree!
this was a very well explained video. thanks so much. this concept is not usually taught that well to be honest so you cleared up a lot of confusion that I Had.
An amazing video, as usual. A clear overview of the content is provided at 13:40 =D
Very useful video, you always make everything so so clear and understandable. Thanks!
Happy to hear that!
Very articulate presentation. Good job!
Thank you kindly!
So helpful! I'm in grad school studying eyewitness testimony and your video helped so much!
Man, this is literally amazing!
Thanks a looooooooooooot!
Good video, I also think of ROC Curve as a visual summary tool to help select the most suitable threshold, without having to go through lots of different confusion matrices.
Great point!
as always, you are very clear and explain everything. i always recommend your videos. thank you!
EXCELLENT VIDEO!! CLEARED MY CONCEPTS BEST.
this is just ... simply amazing!
thanks!
best explanation i have seen so far, Ty
Excellent Sir .Hats off to you 💗❤️❤️
Superb Explanation. Thank you so much Ritvik
Thank you so much, the best teacher out there
Wow, thank you!
Instantly subscribed! Great video, thank you so much.
Thank you SO much for explaining this very confusing concept in such clear terms!
You are doing a really great job man!
I appreciate that!
Masterful lecture, thank you!
My mind is churning with new information!
love the pun
Another video about threshold selection (based on ROC AUC and TPR/FPR or maybe some different methods?) would be a great complement to this!
Great suggestion!
Fantastic explanation!
You explanations are so cool!!
Great as always! Waiting for that SVM video
Coming soon!
yesss, awesome! :D
Excellent explanation
Millions of thanks!
this helped me SO MUCH. Thank you!
Excellent video!
Awesome explanation. Thank you very much
Glad it was helpful!
Great explanation
Amazing video, thank you!
Glad you liked it!
Superb explanation... thanks!
Very good explanation!
Let me just order a pizza and watch all of your other videos as well.
awesome job ...thanks for the great content
Thanks for watching!
Give this man a subscribe!!!
definitely the best !!!
THANK YOU SO MUCH. IT'S VERY WELL EXPLAINED.
Great content. Thank you!
Glad you liked it!
Love your videos!!
Thank you!!
What metric to use for multiclass classification?
Good question. We can still use this ROC curve idea and extend to multiclass. Say we have three classes {0,1,2} and suppose class 1 and class 2 we care about most. Then, we can build one ROC curve for class 0 vs class 1 and one for class 0 vs class 2 and try to optimize both of these.
@@ritvikmathGot it. Thank you so much
Hi Ritvik. Thank you for the great content and really loved your explanation on ROC curve-v helpful. Quick question-so why do we use Precision Recall instead of ROC curve. Do you have video on the same?. I read Precision Recall curve is used more in imbalanced dataset scenario-any inputs here? Thank you:)
This is Naice!!!!!
CANT THANK U MORE 🥰
So great!
Does it assume that a user of our modern will adjust this threshold according to his business scenarios at different time?
Yep that’s right
Thank you
Welcome!
Hi Ritvik, a quick question why dont we set our probability threshold as 0.5 in case of binary classification lets say.
Thanks
Why do we use the area under the curve vs using the minimum distance from the star point (top left corner) to the ROC? Is there any official reason for that?
what do you mean by "Churn "
👏👏👏
👏 👏 👏
awesome
Hi man, I hope you see this comment. Can you please make a video explaining degrees of freedom? and how it's relevant to DS? Thank youu
u salyed it man
💐😍
where is the code? show the code!
Don't you have a Patreon? I owe you money 😂
haha, a good suggestion!
doesn't he look like mini Kash Patel
Quite easy to understand video
Thanks so much @ritvikmath
You're most welcome
Excellent explanation!