Skip to content

AISPLOIT

Back to How Models Learn
FL

Features, Labels, and Targets

Features are what the model sees, and labels are what training compares the prediction against. This room teaches learners how to read a dataset the way a builder would.

35 minHow Models Learneasy95 XP

Listen to hear this room section by section.

Key Ideas

Work through these sections in order. Each one builds the mental model you need before the checkpoint questions will feel easy.

In supervised learning, features are the inputs the model is allowed to use when producing a prediction. Labels are the target answers the training process compares those predictions against. If you blur those two roles together, you lose track of what the system can know at prediction time versus what it is trying to predict.

Suppose you are building a churn model. Recent usage, number of support tickets, subscription age, and region might all be features. Whether the customer churned in the next month might be the label. The model reads the features and tries to produce the label.

A useful builder question is: if I deploy this model tomorrow, which fields are available before the prediction is made? Those are the candidate features. Then ask: which field represents the answer I want the model to learn to predict? That is usually the label.

This distinction sounds simple, but it is one of the most important habits in the whole path. Many later problems, including leakage and bad evaluation, begin with confusing the target for an input or treating post-outcome information as if it were available ahead of time.

You've opened 1 of 4 sections. Once the ideas feel clear, move into the checkpoint block below.

Check Your Understanding

These checkpoints reinforce the lesson you just read. If one feels fuzzy, reopen the relevant section above before trying again.

3 checkpoints
1

Task 1

Choose the valid features

Select the fields that could plausibly be used as inputs in a home-value model.

Which fields are plausible features for predicting a home's value?

2

Task 2

Classify the fields

Match each field to the role it plays in a churn dataset.

For each field, choose whether it acts like a feature, a label, or an identifier.

Daily active minutes over the last 30 days

Whether the customer churned in the next month

Randomly assigned customer record ID

3

Task 3

Explain feature versus label

Write a short explanation of the difference between what the model reads and what the training process scores.

In one or two sentences, what is the difference between a feature and a label?

Ready To Move On?

Up next: What Training Actually Changes

Back to PathContinue to Next Room