Class 11 Machine Learning Logistic regression in Python

You can find the slides and notebook on my GitHub repository for the course:
Topics covered: Logistic Regression in Python, Classification performance metrics (accuracy, precision, recall, f1 score), ROC and AUC, Precision Recall trade off and f1 score.
- Example: predicting the probability of default and maximizing lender's profit function.

Here is a Crash course in machine learning concepts:

Here is a detailed course on machine learning and its applications in finance:

[clatacis]

I liked the real world example where we were able to see the profit function and how different thresholds would change the maximum profit. It was exciting to me to see how data can be used for so many applications besides precision, recall, and etc.

[reginemartschiske]

Dear Pedram, in the video above you said logistic regression has no hyperparameters, but what about 'C' and 'penality'? You used cross_val_score to estimate the F1-Score in the test set. But what does it mean if the F1-Score is severely underestimated or overestimated? Do we always use cross_val_score to estimate a Score in the test set? The other way ist to fit a model and calculate the F1-Score in the test set via f1_score(y_test, y_test_predict). Why do you prefer so estimate the Score via cross_val_score - is it because of reducing computing time?
Thanks a lot for your great video lections!!!!

[davidjung]

You may have mentioned this, but in using Logistic regression are we able to add a third dimension/classification? For example, if we want to classify probability of default by income AND monthly credit payment? Does this help improve accuracy or does it depend?

[Sarahchamberlain]

Is there an industry standard boundary between imbalanced and balanced? Would a 50/50 split in the target variable be ideal?