Logistic Regression with Imbalanced data: A Geometric View

[@geofisico2007]

This is, BY FAR, the best explanation on how imbalanced datasets damage data modeling results. It can be extended to other algorithms, such as SVM. Cheers from Brazil and thanks a tons, sir!

[@AppliedAICourse]

Yes, similar argument can be made for Linear SVMs also. Happy that you liked it.

[@abhayram375]

Tons of thanks for this brilliant explanation Sir !!!

[@ZshaanKhan]

Thanks for the explanation and we have that kinda assignment for all the linear models to understand well how the imbalance impacts the models and how the hyperparameter helps. Thankyou Team for the brilliant stuffs you have. ✌🏼✌🏼

[@AppliedAICourse]

Yes, we have this as part of one of our assignments so that students learn to appreciate these real world cases. Here, we tried to explain it visually.

[@kaustuvdash4422]

So sir why to applied sigmoid function in imbalanced data set..... Use tanh function so that missclassfied values will get a value of -1 rather than 0...so in case of imbalanced data set also it will work fine..... Correct me if I am wrong?

[@sagarverma13]

Perfect Explanation sir.. Hats off...

[@balajichippada]

Perfect explanation as always!!

[@surajshivakumar5124]

SVM will not have the same problem as it is only dependent on support vectors right?

[@kumarabhishek5652]

great explanation sir.

[@kushshri05]

Haven't found this kind of explanation anywhere 🙂

[@barunbodhak720]

Fantastic explanation sir.

[@fahadreda3060]

Very informative video

[@shashidharyalagi407]

Perfect explanation

[@CRTagadiya]

but if we use regularization then pi(1) will be selected right?

[@devallaeshwar3660]

How to measure the performance of the model for such imbalanced data sets??

[@AppliedAICourse]

The exact metric depends on the problem being solved. But as an example, you can use metrics like precision and recall for both the majority and minority class to get a sense of how well the model is working on each of the classes.

[@revarevanth1800]

How the equation came yWx could you please explain and how that 0.8 is assumed could you please tell those tiny details

[@AppliedAICourse]

The derivation of the logistic regression equations from scratch takes a couple of hours of videos. The 0.8 is taken as one example value for the output of the sigmoid function used in Logistic Regression. This was an answer given to a student’s question and assumes that the viewer knows about Logistic regression and the underlying geometric and mathematical details.

[@thedanglingpointer60]

Why not upload this with the coursework.Would be helpful to us

[@AppliedAICourse]

We added it to the comments section as it was a student query. We encourage our students to read through the top 2-3 voted comments under each video to gain a deeper understanding. We also added it at end of Module 3 here: www.appliedaicourse.com/lecture/11/applied-machine-learning-online-course/4236/logistic-regression-with-imbalanced-data-a-geometric-view/3/module-3-foundations-of-natural-language-processing-and-machine-learning

[@yudiutkarsh5173]

Sir can u post eda data loading vedio

[@arun_sain]

👌👌👌