Logistic Regression Explainer
Logistic regression is a supervised classification model. For binary outcomes, it computes a weighted sum of input features and passes that score through the logistic function, producing a value between zero and one.
Training selects weights that maximize likelihood or equivalently minimize a log-loss objective. Regularization limits coefficient size, helping control overfitting and instability when features are numerous or correlated.
A decision threshold converts the probability score into a class label. Changing that threshold changes the balance between false positives and false negatives, so it should reflect the application rather than default automatically to one-half.
The model is valued for speed, strong baselines, and coefficients that can be interpreted under careful assumptions. It still represents a linear boundary in the supplied feature space, and coefficient meaning can be distorted by scaling, interactions, confounding, or poor calibration.
Logistic regression is a linear classification model that converts a weighted feature score into a class probability.
Key terms include linear score, logistic function, probability, log loss, coefficient, regularization, decision threshold, odds and calibration.
The logistic function maps any weighted feature score into the zero-to-one interval used for binary probability estimates.
Multiclass extensions compare several class scores through multinomial or one-versus-rest formulations.
A threshold determines which probability scores trigger the positive class and therefore changes error tradeoffs.
Teams should choose it with validation data, operating costs, class prevalence, and calibration in mind.
Explore more "Explainers"
Discover additional explainers across politics, science, business, technology, and other fields. Each explainer breaks down a complex idea into clear, everyday language—helping you better understand how major concepts, systems, and debates shape the world around us.
