← AI Terminology
Logistic Regression
Logistic regression is a linear classification model that predicts class probabilities via a logistic/sigmoid of a weighted feature sum, trained with log loss.
It is the baseline classifier every practitioner should beat.
It is the baseline classifier every practitioner should beat.
Why It Matters in AI
Before deep nets, logistic regression ran much of industry ML — and still does for simple tabular problems. It is interpretable, fast, and the conceptual base for neural binary heads and calibration discussions.
Key Points
| Aspect | Description |
|---|---|
| Cons | Linear decision boundary unless features engineered |
| Loss | Binary (or multinomial) cross-entropy |
| Pros | Simple, strong baseline, coefficients interpretable |
| Model | P(y=1 |
| Tools | scikit-learn LogisticRegression |
| Related | Linear regression, softmax, GLMs |
Simple Analogy
Drawing a straight fence in feature space and reading the distance to the fence as a probability of being on the “yes” side.
Common Usage Examples
sklearn.linear_model.LogisticRegression- Baseline before XGBoost
- Coefficient inspection for drivers
- Multinomial logistic for multi-class
Summary
In short: Logistic regression turns a linear score into a class probability — the classic, interpretable classification baseline.