Logistic regression is the default model for binary classification, and it is one of the most heavily tested models in a quant interview because it sits exactly at the intersection of linear models, probability, and optimization. It answers a yes-or-no question with a calibrated probability: will this trade be profitable, will this loan default, is this order toxic? This lesson shows how to turn a linear score into a probability with the sigmoid, why the loss is cross-entropy and not squared error, how the model is fit by gradient descent, and how to read the coefficients as odds ratios. We work two examples by hand and close with the pitfalls and the threads into the rest of the course.
Table of Contents
Start Your Interview Preparations!
Already have an account? Log in!