Feature, in machine learning, is an individual measurable property or characteristic of the data that a model uses to make predictions. Features are the inputs the model learns from.
What it means in plain English
When a model looks at data, it does not see the whole thing at once — it sees a set of features. Each feature is one piece of information about an example. Choosing which features to give a model, and how to represent them, has a big effect on how well it can learn. The right features make a pattern easy to spot; poor ones can hide it.
For structured data, features are often the columns of a table; for images or text, they are learned automatically by the model.
A simple example
To predict whether a customer will cancel a subscription, useful features might include how long they have been a customer, how often they log in, and how many support tickets they have raised. Each is a feature the model weighs.
Why it matters
Features are how data is described to a model, and choosing good ones is one of the most important parts of building an effective machine learning system. Much of a data scientist’s craft lies in identifying and shaping the features that let a model learn well.
Related terms
- Feature Engineering — the craft of creating good features.
- Dataset — the collection features are drawn from.
- Model — what uses features to predict.
Frequently asked questions
What is a feature in machine learning?
A feature is an individual measurable property or input variable the model uses to make predictions — for example, a house’s size, location, and age when predicting its price.
How do features affect a model?
The choice and quality of features strongly affect performance — good, relevant features make a model’s job much easier, which is why feature engineering matters.
