logistic regression

Gold definitionUpdated Apr 2, 2026

Definition

Logistic regression is a statistical model used for binary classification, predicting the probability of an event occurring by fitting data to a sigmoid function. It's a lightweight, interpretable classifier often employed in scenarios like rare-event prediction to mitigate bias in imbalanced datasets.

At a glance

Executive summary

Logistic regression is a simple yet powerful statistical tool that predicts the probability of a yes/no outcome, like whether a customer will churn or if a disease is present. It's valued for its clarity and efficiency, especially when dealing with situations where one outcome is much rarer than the other, helping to make more accurate predictions in critical areas.

TL;DR

A mathematical model that predicts the probability of a binary outcome (like yes/no) based on input factors, often used for its simplicity and effectiveness in predicting rare events.

Key points

Uses a sigmoid function to model the probability of a binary outcome based on a linear combination of features.
Provides interpretable probabilistic predictions for binary classification, effectively handling class imbalance in rare-event prediction.
Used by researchers and engineers in healthcare, finance, marketing, and reliability engineering for classification and risk assessment.
Simpler and more interpretable than complex neural networks, but less capable of capturing highly non-linear relationships.
Often used as a baseline or as a component in hybrid systems to add interpretability or correct biases in more complex models.

Use cases

Healthcare: Predicting the likelihood of a patient developing a specific disease (e.g., diabetes, heart disease) based on medical history.
Finance: Assessing credit risk for loan applicants, predicting the probability of loan default or fraudulent transactions.
Marketing: Predicting customer churn (whether a customer will stop using a service) to enable targeted retention efforts.
Rare-event prediction: Identifying potential equipment failures in reliability engineering or safety incidents in aviation.
Customer Support: Classifying customer inquiries into categories to prioritize urgent requests or identify common issues.

Also known as

Binary Logistic Regression, Multinomial Logistic Regression, Ordinal Logistic Regression, Logit Model