Purchase-Intent Classifier Notebook Under Weighted F1
Problem statement
How would you build a purchase-intent classifier in a timed notebook assessment? You receive labeled store sessions and an unlabeled test file, with inconsistent values, missing data and coded categories. Purchases are uncommon, and scoring uses weighted F1. Deliver a documented notebook and a submission with an id and binary outcome for every test session.
Explain cleaning, weighted F1 and the complete notebook workflow.
Show hint
For each odd value, ask whether it is a real measurement or a code or unit bug that a data pipeline introduced.
Relevant roles: ML Engineer, Data Scientist and Applied Scientist.