Rows and Columns: The Atoms of Data

Pick almost any dataset in the world, from student grades to the famous Iris flowers dataset from UCI, and it will probably look like a table. Rows going down, columns going across. This simple shape is the workhorse of data science.

Example: the Titanic dataset

Kaggle has a beginner-famous dataset about the Titanic. Each row is one passenger. Each column is one thing we know about that passenger: name, age, ticket class, whether they survived.

Vocabulary you will hear

• Row = record = example = instance = sample
• Column = feature = attribute = variable = field
• Cell = value at one row and one column

When tables go wrong

• Two things stuffed into one row
• A column that means different things in different rows
• Empty cells that should not be empty
• Duplicate rows of the same person

The big idea: rows are examples, columns are properties. This simple mental model will carry you from your first spreadsheet to training a million-row model.