145 - Confusion matrix, ROC and AUC in machine learning

[@fadilyassin4661]

I think you are one of the best tutors I ever seen on deeplearning you are truley are this is so professional and thats what every tuting must be thank you

[@DigitalSreeni]

Thank you for your kind comments.

[@bencerepasi5352]

Hi Sreeni! I've found your channel during my time writing bachelor thesis about GMM in regards of pixel-wise semantic segmentation. I've learned a lot from your channel in the beginning. Now I'm working on the industrial use of image classification with Neural Networks and I have to say, I'm still going back to your videos. The quality is amazing and you are great at explaining complex methods in concise and understandable manner. Cheers from Austria!

[@DigitalSreeni]

I am glad you find my videos to be useful. I often go back to my old videos to refresh on topics I haven't explored in a while. Salzburg is one of my favorite places to visit, not just in Austria but in the entire world. I love Altstadt and the view of Salzach River from the top of Fortress Hohensalzburg is breathtaking.

[@tanaydasbarman1600]

Thanks!🤩

[@roger_island90]

I can't walk away without subscribing to your channel. You are the best and I dove out my heart for you sir.

[@DrRubidium]

Yes, you are right. I love this channel. This topic was quite clear to me, but you added even more graphical insight and now it is crystal clear. Thank you.

[@gauravkanu4486]

I am so glad to watch this video, your explanations are so precise and amazing. I highly doubt though how come you have only 27k subscribers? ♥♥♥♥♥☻

[@VivekTR]

Is the confusion matrix in 4:34 correct? Uinfected correctly being labelled as uninfected is true negative is it not?

[@muskanvij4811]

Reallly good ! Thank you so much. Best video on u tube on confusion matrix.

[@DigitalSreeni]

Glad you liked it!

[@aylinwagner8047]

Great, finally I understood this !!!

[@Jihaoui]

I watched many of your videos, they are all very heplful and greats. Please if possible to provide mor videos about tabular data.

[@caiyu538]

Great lectures. I follow up your series. I leaned a lot.

[@ElfengourAbdelhakGIS]

This video is really well explaind, keep ip the good work.

[@DigitalSreeni]

Thanks a lot!

[@ML-DS-AI-Projects]

Thanks for the detailed explanation. A good model balances the TP,FP,TN,FN? Or it balances the sensitivity or specificity metrics?Need clarification.

[@saikrishnavadali05]

Awesome Video... DigitalSreeni is the best!!!

[@aditiarora2128]

sir how to plot confusion matrix and roc when we are using image augmenatation but not X_train and Y_train,Y_Pred,Y_test....plz help

[@DesertWolf90]

it's possible to calculate the Ideal threshold for segmentation tasks or its only for classification tasks ?

[@Rider-jn6zh]

Did not get how to choose optimal threshold

[@junaidlatif2881]

Roc only for binary classification while confusion matrix could be used for both binary and multi class classification! 😍

[@honeyrulesintheworld]

can you tell me how to get confusion matrix for image retrival using CNN?

[@solacing]

How can we implement this (ROC-AUC) in multiclass image classification?

[@padma386]

Nice explanation

[@jalpeshdadania562]

Wow. Great explanation. Thank you so Much

[@sabaal-jalal3710]

ValueError: Layer model expects 1 input(s), but it received 617 input tensors. Inputs received, I got this error when I run the code: " from sklearn.metrics import confusion_matrix y_pred = (model.predict(X_test)>= mythreshold).astype(int) cm=confusion_matrix(y_test, y_pred) print(cm) " on my model the test set is 617 images, how to fix this problem sir?

[@alroygama6166]

But what can we infer between a threshold of 0.5 vs 0.7 which is better. What can you say about these

[@haoduong6565]

Hi, when you calculate confusion matrix, if you don't choose 0.5, what is the default value used to calculate it? Thanks!

[@Bryan-eg7si]

Excellent video. Very clear !

[@helloworld_hereiam1402]

Awesome presentation. Could you please do a video on confusion matrix, AUC curves for multilabel classification using xtest and ytest? Thanks.

[@DigitalSreeni]

Will try

[@helloworld_hereiam1402]

@@DigitalSreeni many thanks

[@Juliodonadello]

Excellent video man

[@hani97tn]

Thank you! So clear and useful. but there is a lot less of cell images to train with in the folder.

[@junaidlatif2881]

Sir how to use at RF?

[@libby3121]

Thanks for all your tutorials. Can you pls tell why there is a need to ravel the predictions ? ()

[@Mordi567hj]

un-transform the scaling

[@mohamedbachiri7891]

useful information Thank you 👍

[@DigitalSreeni]

Welcome 😊

[@izharl]

Thank you! In wich situations is more relevant to use confusion matrix and in which ROC?

[@DigitalSreeni]

Confusion matrix tells you the accuracy for each class and ROC helps you assess the model performance, especially in comparison to other models. So if you are plying with many models to see which one is appropriate then use ROC metric.

[@eventhatsme]

Great explanation and demonstration! Is the threshold test vs confusion matrix only available for certain classifiers or only for binary classification? I have tried adapting this to multiclass randomforests and that classifier appears to only output class numbers and not per class probabilities.

[@junaidlatif2881]

Best explanation... ❤

[@helloyes4795]

hello Sir! thank you very much for your good work in helping us to get acquired about everything needed for a microscopist. I have a problem, which ways can you advise me to pass through in order to count colonies of A PETRI DISH ( colony counting for petri dish) thank you for your considerations?

[@DigitalSreeni]

We did this task an year ago as part of my work. Here is the link to a pre-configured workflow. (It is a free platfrom, don't worry). www.apeer.com/app/workflows/detail/Colony-Forming-Unit/8cfed555-50ce-4eda-be1b-b3202c1e5e9f If you want to write your own code then there are many way but the best one that works most of the time would be to use some sort of machine learning. I recommend starting with traditional machine learning, please watch my videos on this topic (from video 60 to 70 or so). If you want to label your images for machine learning then you can do it here (again, free): www.apeer.com/annotate

[@helloyes4795]

@@DigitalSreeni thank you very much Sir

[@abderrahmaneherbadji5478]

The video is very helpful, thanks to you

[@DigitalSreeni]

Glad to hear that

[@holthuizenoemoet591]

uninfected being 1 is really offsetting. The rest of the video is great

[@DigitalSreeni]

Yes, we all can use some mind exercise sometimes :)

[@benjaminschnepf7362]

your videos are great!

[@pedroramos7814]

very good. thank you!

[@DigitalSreeni]

Glad you liked it!

[@helplearnhome9542]

thanks a lot!

[@DigitalSreeni]

You're welcome!