Classification

Classification is a predictive task that assigns observations to discrete categories. A classification model learns a rule from labeled examples or other specified information and applies it to new cases.

Training set: A collection of examples used to fit a predictive model. Classification algorithms use these examples to learn how observations map to categories.
Supervised learning: A machine-learning setting in which models learn from examples paired with known outcomes. Classification is one of the principal predictive tasks within supervised learning.
Spam filtering: The process of identifying and separating unsolicited or unwanted messages from other communications. Email systems classify messages into spam and non-spam categories.
Regression analysis: A method for modeling relationships between variables, often to predict a continuous outcome. Regression predicts numeric values rather than discrete categories.
Feature: A measurable property or input variable used to represent an observation in a model. Features give a classifier the information it uses to distinguish categories.
Class label: A discrete value identifying the category assigned to an example. Class labels define the possible outcomes a classifier predicts.
Medical diagnosis: The process of identifying a disease or condition from a person’s signs, symptoms, and test results. Diagnostic systems classify clinical observations into possible conditions.
Clustering: A method that groups observations by similarity without requiring predefined category labels. Unlike supervised classification, clustering discovers groups rather than learning known classes.
Decision boundary: A surface separating regions assigned to different classes in a model’s input space. It shows how a classifier divides observations among predicted categories.
Binary classification: A classification task with exactly two possible categories. It is the simplest setting for studying category prediction.

Linked from 9 pages