In the previous lesson you learned that a model learns from examples instead of being told exact rules. This lesson is about those examples. Every example is a small package of facts, and a pile of examples stored in an organized way is called a dataset. The model reads the dataset over and over, notices patterns, and turns those patterns into the ability to make predictions. If the examples are correct, plentiful and fair, the model can learn something true about the world. If the examples are wrong, too few, or lopsided, the model faithfully learns the wrong thing. Data quality is not a boring detail — it is the single biggest factor in whether a machine-learning project works.

Learning goals — after this lesson you will be able to:

• Describe a dataset as a table of rows (examples) and columns (features), and point to the label column.
• Name the main feature types — numeric, categorical, ordinal, text and date — and give an example of each.
• List what makes data good and the five common problems that break models.
• Explain "garbage in, garbage out" with a real worked example.
• Say where data comes from and what responsibilities come with collecting it.
• Read and inspect a CSV file in Python with a few lines of pandas.