AI glossary · Beginner · Classic ML, recsys & search
Regression analysis
Regression analysis is a method that finds the relationship between several input values and an outcome value, and then predicts a continuous number such as a house price, sales figure or temperature when a new input comes in. It is one of the most basic forms of supervised learning.
About this term
- Level
- Beginner · Core ideas for first-time readers
- Category
- Basics
- Added
- 2026-10-08
How it works
From past data, inputs (e.g. floor area, number of rooms, distance to a station) are collected together with the correct numeric answer (e.g. house price). The model builds a formula that multiplies each input by a weight and adds them up, then adjusts the weights so that the difference between predicted and actual values (the error) is as small as possible. The simplest form, linear regression, treats this relationship as a straight line (or plane); more complex relationships can be handled by solving the regression problem with models such as decision trees, gradient boosting or neural networks. Although its name contains 'regression', logistic regression, which outputs a 'yes/no' probability, is actually used for classification.
Regression vs. classification
RegressionHow much?
- Predicts a continuous number
- e.g. house price, sales, temperature
- Evaluated by size of error
ClassificationWhich one?
- Picks one of a fixed set of categories
- e.g. spam or not, product type
- Evaluated by share of correct answers, etc.
공통 Both are supervised learning, trained on data pairing inputs with correct answers
Why it matters
Most tasks that ask 'how much', such as demand forecasting, pricing and risk scoring, are regression problems. Linear regression also has the advantage of being easy to explain, because its weights directly show how much each input affects the outcome. That is why a regression model is often built first as a baseline before creating complex AI models. It is also the best starting point for learning basic machine learning concepts such as loss functions, gradient descent and overfitting.
Things to keep in mind
What regression finds is a 'relationship of moving together', not cause and effect. Ice cream sales and swimming accidents may rise together without one causing the other. Inputs outside the range of the training data can lead to badly wrong predictions, and adding too many input variables causes overfitting. A common problem is a few extreme outliers pulling the entire line toward them.
Example
Real estate platforms show a home's estimated market price from information such as floor area, floor level, location and nearby facilities, and retail companies forecast next week's sales from weather, day of the week and whether there is a promotion, so they can prepare inventory. In both cases the outcome is a continuous number, so they are treated as regression problems.