Imbalanced Classification Pipeline with SMOTENC
Problem statement
How would you build an honest fraud-classification pipeline for orders? Fraud is rare, and the table mixes numeric and categorical features with missing values. You need to compare resampling with class weights, tune models and choose a useful decision threshold.
Build and evaluate the pipeline, explaining how you prevent leakage.
Show hint
Which steps in this pipeline learn something from the data, and on which rows is each allowed to learn it?
Relevant roles: ML Engineer, Data Scientist and Applied Scientist.