Clear thinking about intelligent tools.
Class weights alter a classifier’s training penalties. Use a small numerical example to see why they do not change the real-world frequency of a class.
Read permutation importance cautiously when model inputs carry overlapping information, and distinguish a small score drop from proof that a feature is useless.
Stratification can preserve class proportions in a train-test split. It does not automatically balance every feature or prevent data leakage.
Macro and micro averaging answer different questions about classifier performance. Use a small example to see why class balance matters.
See how a classifier’s decision threshold changes false alarms and missed cases, using a small worked example and a fixed evaluation set.
Distinguish a model’s confidence from demonstrated reliability, using calibration checks on representative data before treating a score as a probability.
Spot machine-learning data leakage by checking when information became available and keeping test data out of model preparation.
Build a computer-vision evaluation around expected conditions and costly mistakes, with a worked example that separates test images from training material.