Predict a binary outcome (0 / 1). Linear combo of features is squashed by σ(z) = 1/(1+e⁻ᶻ) into a probability between 0 and 1. Threshold (default 0.5) decides class.
Coefficients—
Threshold0.50
Train accuracy—
Decision boundary in feature space
Sigmoid σ(z) — squashes z into P(y=1)
0.50
Probability → odds → log-odds: the linear part z = β₀ + β·x is the log-odds. Sigmoid converts that back to a probability. Slide the threshold to trade off false positives vs false negatives — go lower to catch more positives at the cost of more false alarms.