Swap In the Typical Value

A random forest has no coefficients, but you can still poke one prediction: replace one of the row's values with the typical (median) training value, and see how far the chance of benign moves.

Task: write biggest_swing(test_index, columns).

• Load load_breast_cancer(return_X_y=True, as_frame=True) and split with train_test_split(X, y, test_size=0.25, random_state=0, stratify=y).
• Fit RandomForestClassifier(n_estimators=100, random_state=0) on the training rows.
• Take row = X_test.iloc[[test_index]]. Its base chance is predict_proba(row)[0, 1].
• For each name in columns: copy the row, set that column to X_train[name].median(), and compute base - new_chance.
• Return a dict:
  • "base": the base chance, rounded to 2 decimal places
  • "changes": a dict from each name to its change, rounded to 2 decimal places
  • "biggest": the name whose change is largest in size, ignoring the sign

All numbers are plain floats. No two changes tie in size.