What are Features and Labels in AI and Machine Learning?

Features are the independent factors in machine learning and are synthesized to create an outcome (prediction) called as labels. Features are an individual measurable factor while the value of labels depend on the features.

In machine learning problems, features are often represented by “x” and labels as “y“.

Characteristics of Features and Labels

Features

Features are those factors in AI-ML that are independent, measurable and whose value does not depend on any other factor.

1. They are independently measurable, such as number of room in a house, the existence of cancer cells in a diagnosed patient, the length of a certain car, the number of engines in a ship.
2. They are directly affected by bias. For example, an Asian person may show different behavior towards the need of discipline than an European. This would later impact an AI analysis of “Impact of Strict Parents in Child’s Future Income”.
3. Exist in variety of forms. Features may exist in extreme forms. For example, the type of cancer prevailing in the USA could be completely different than the types of cancer in the sub-Saharan Africa.

Labels

Labels are those extracted results that are obtained by manipulating features though a mathematical equation. They derive their value from features.

Consider these

Factors That Affect Features and Labels

Features are affected by several things while labels are only affected by those features. Hence, if there is a bias, it will not only create an inaccurate feature but also negatively impact the label and hence corrupting the contribution of other features as well.

1. Clean Data Set
2. Zero Bias
3. Accurate Features Measurement

Examples of Features and Labels

Example 1: Identifying Cancer Cells in the Human Body

In a hypothetical cancer cell study, the aim is to study if there is a correlation of lifestyle factors with cancer or not.

Features:

1. Smoking
2. Drinking alcohol
3. Usage of recreational drugs

Here, the Label is the identification of an existing cancer cell.

Example 2: Predicting House Prices in Los Angeles

In a study to find which factor has the highest contribution in house prices in Los Angeles.

Features

1. Number of rooms.
2. Tiled, wooden or stone floor.
3. Garage size.
4. Pool Size.
5. Floors in the house.

Here, the label would be the price of the house.

Example 3: Classifying Consumers in Individual Groups

In a study to classify consumer behavior, i.e., to classify them as impulsive spenders or careful planners, the features and labels would be:

Features

• Whether they prefer branded wine
• Whether they use credit cards or cash
• Whether they buy a gadget on the launch week
• Whether they get all the latest iPhones

Here, Labels would be their classification.