Feature

Definition

A feature is one measurable property of an input.

Example:

house:
rooms = 3
size = 1200
age = 20

Features:

rooms
size
age

Feature = one input property used by a model.

---

Feature As Vector Component

When data is represented as a vector, each component can be a feature.

Example:

input = [3, 1200, 20]

could mean:

feature 1 = rooms
feature 2 = size
feature 3 = age

Features are the meanings attached to vector components.

---

Feature Consistency

Feature order must stay consistent.

Good:

[rooms, size, age]
[3, 1200, 20]
[4, 1500, 10]

Bad:

[rooms, size, age]
[3, 1200, 20]
[10, 4, 1500]

The second row uses a different order.

That breaks the representation.

Feature positions must keep the same meaning across examples.

---

Feature Values

A feature has a value for each input.

Example:

feature: size
value: 1200

In vector form:

[3, 1200, 20]

1200 is the value of the size feature for that input.

A feature names the property. A feature value gives its value for one input.

---

Features And Classification

A classifier uses features to make a decision.

Example:

input features = [hours studied, attendance]
class = pass or fail

The classifier does not see the real-world object directly.

It sees the feature representation.

Classification depends on how inputs are represented as features.

---

Good Features

A useful feature helps separate classes.

Example:

feature: number of missed payments
class: risky or not risky

This feature may carry useful information.

A weak feature may not help.

Example:

feature: random ID number

If it has no meaningful relation to the class, it may not help classification.

Features matter because they shape what the classifier can use.

---

Feature Scale

Feature scale can affect linear models.

Example:

rooms = 3
price = 300000

Large-scale features can dominate a weighted sum if not handled carefully.

This is why normalization can matter.

Feature scale affects numerical comparison and weighted sums.

---

Final Summary

• A feature is one measurable property of an input.
• In vectors, features are component meanings.
• Feature order must be consistent.
• A feature value is the value for one input.
• Classification depends on feature representation.
• Useful features help separate classes.
• Feature scale can affect linear models.