Episode 4: Simple and Basic Binary Classification Metrics

In this episode of applied machine learning series, we learn about the most basic, but important, #binary #classification metrics and enhance our mlframework with them. We create a classification metrics class which can be expanded to any metric in future.

Follow me on:

[NitinGupta]

This series is fantastic, Abhishek! Could you also do a video on ensemble machine learning techniques (stacking models)? Thanks!

[allieubisse]

One day I'll be a grandmaster like you! And give back to the community by sharing my knowledge to the young stars.

You inspire me!🥇

[lalitgarg]

Now i become fan of your handwriting also boss...you are doing cool stuff....thanks for doing all these things...we all are enjoying your videos and content specially framework part on how to approach any problem and simplifying the code by standardizing the things and get rid of redundant works and efforts again and again.


I am waiting for your first book also eagerly on ML Framework....

[AlexeyMatushevsky]

13:18 Actually AUC = 0 is very good model, we just need to flip the predictions 1-pred; right?

[ajinkyagangurde]

Thanks for this informative video! What metric should be used if I want higher lift in top deciles. F1 or AUC? and why? My data has 25% 1s and I am interested in both 1s and 0s.

[debarchanbasu]

Hi Abhishek da! I didn't understand the part where you say that AUC value represents the probability that a +ve data point "ranks higher" than a negative data point. Could you please elaborate what is meant by "ranks higher" ?

[uthamkanth]

awesome., looking forward for more advanced metrics:)

[Ahmed-ozfn]

Hi Abhishek!
Pls, what is the model of handwriting graphic tablets are you using.
thanks.

[allieubisse]

Please help calculating f1_score for multi class text classification (sentiment analysis)

[FREELEARNING]

Thank you for this fantastic series of Applied ML. It would be helpful if you could make a video about Loss Functions, how to create custom loss functions.

[anuraglahon]

Hey can you suggest me a good data set for regression and classification for my final semester project

[vrajbhatt]

Sir I need your support for affinity propagation... having some issues in my project

[hosseinmortazavi]

Hi Abhishek,
what metric do you suggest for a binary class which is imbalanced like a churn model with 96% negative(not churn) vs 4% churned user?

[ChandraKanth]

Hello Sir, Thanks for this material. Can you do the same one for Regression?

[adarshnamdev]

Really wonderful video on the performance metrics Abhishek, Thanks for creating one!
I have a doubt of which I couldn't find a proper answer on internet. Why AUC is only plotted using TPR and FPR on Y and X- axis respectively?

[noisiplayer]

When recall = TP / (TP + FN), precision = TP / (TP + FP) and F1 = (2 * recall * precision) / recall + precision should´t be F1 = (2 * TP * TP) / (2 * TP + FP + FN) which is (2 * TP squared 2) / (2 * TP + FP + FN) ? You wrote just 2* TP at 7:52. If not I am sorry :D