Unit 2 · Where the effort is
The unit that decides whether a project succeeds. A model can only learn what its data contains, including the data errors and the historical bias.
What a strong answer looks like
A strong Unit 2 answer says what was done about missing values and encoding, and justifies each choice rather than reporting it.
Know
A model reproduces the relationships in its training data, including systematic errors. No algorithm recovers information the data never held.
Apply
Inspect shape, types, missing counts, and class balance before modelling.
Watch out
Assuming clean data. Real datasets have gaps, duplicates, and wrong types by default.
Study move
Profile a dataset and list every problem you find before touching a model.
Know
Leakage is a feature that would not exist at prediction time, or that encodes the answer. It produces excellent scores and a useless model.
Apply
For each feature, ask whether it is available at the moment the prediction must be made.
Watch out
A cancellation date in a churn model, which is present only for customers who already churned.
Study move
Mark each column of a dataset as feature, label, or leaking, with a reason.
Know
Integer codes imply an order categories may not have, so unordered categories are one-hot encoded. Distance-based algorithms need features on comparable scales.
Apply
One-hot unordered categories; scale features before any distance-based method.
Watch out
Fitting a scaler before splitting, which lets test statistics influence the transformation.
Study move
State the correct order of split, fit, and transform, and what breaks if reordered.
Know
Every handling choice changes the result. Dropping rows can remove exactly the group under study; imputing with a constant distorts a distribution.
Apply
Choose a documented strategy and report it alongside the result.
Watch out
Replacing missing values with zero without checking whether zero is a meaningful value in that column.
Study move
Report missing counts per column and justify one handling choice.
Connections and techniques that receive extra attention in this unit.
Related topics some schools attach to this unit and others leave out. Covered on request rather than assumed.
Modeling drill
Given a dataset predicting exam success, list usable features, name one leaking column, and state your missing-value strategy with a reason.
