2 Datasets and Features Free Online FlashCards

What is a dataset?

A collection of examples used to develop or apply a machine-learning model.

What is a key limitation of a proxy label?

A proxy label approximates the intended outcome, but may not match the real goal.

Why does a dataset need representative samples?

If samples do not reflect the cases the model will encounter, its predictions may not work well in that real-world setting.

When should a prediction feature be available?

Features should be available when the prediction is made; using information that becomes known only afterward can mislead evaluation.

What is the first step in preparing data for modeling?

Define what one sample represents and which label the model should predict.

What is a sample?

One case that a model learns from or makes a prediction about; in a table, it is often a row.

What is a feature?

An input attribute that may help a model make a prediction; in a table, features are often columns.

What is a label?

The known outcome that a model is intended to predict.

What makes a sample labeled in supervised learning?

A sample that contains both input features and a known label.

What consistency issues should be checked in a dataset?

Check that formats, units, and category names are consistent across the data.

What should data validation inspect?

Check row and column meanings, types, units, plausible ranges, missing values, duplicates, and label consistency.

How should missing or erroneous values be handled?

Correct errors when possible; otherwise, decide whether to remove affected samples or fill missing values appropriately. Record important choices.