Phase 3 · Machine learning · Models
Logistic regression
Predict probabilities for yes/no outcomes.
Open this lesson Create a free account
Why it matters
Most business predictions are yes/no: will this customer churn, will this lead convert, is this review negative. Logistic regression answers them with a probability, trains fast, and its coefficients are readable. It's the standard first model for any classification problem, and its structure (a linear score through a squashing function, trained with log loss) is exactly one neuron of a neural network.
What you'll be able to do
- Turn a linear score into a probability with the sigmoid
- Train logistic regression with log loss and gradient descent, from scratch
- Read coefficients as changes in log-odds and odds ratios
- Choose a decision threshold for turning probabilities into actions
Before this lesson
Next in this phase: Multiclass classification