Module 2 lesson

Data Types and Column Roles

Unit ID: ML-M02-U02 Estimated active time: 25-30 minutes

Type is not the same as role

A column has a data type, such as text, number, date, or category.

A column also has a role, such as target, feature, metadata, or excluded.

These are different questions.

For example, final_quiz_score is numeric. But it is not a valid feature for a day-3 prediction because it is recorded later.

Common ML column types

Type	Example	Note
Numeric	`day3_minutes_active`	May need missing-value handling
Categorical	`python_preparation`	May need encoding later
Date/time	`started_module1_date`	Useful for timing checks
Binary	`completed_module1_by_day10`	Often a target
Identifier	`record_id`	Usually metadata, not a feature

Inspect with pandas

Use:

df.info()
df.head()
df.dtypes

These commands do not tell you everything, but they help you start.

object columns in pandas often contain text or categories. You must inspect the values before deciding how to use them.

Practice

For each column in the Module 2 dataset, fill:

Column	Type	Role
`record_id`	text	metadata
`completed_module1_by_day10`	binary	target
`final_quiz_score`	numeric	exclude future information

Then complete the remaining columns.

Takeaway

Column type tells you what kind of value is stored. Column role tells you how the value may be used. A numeric column can still be unsafe for modelling.