Three Ways to Prepare

Does scaling help, and which scaler helps most? Compare three versions of the same model on the same split.

Task: write compare_preparation(name, k). name is "wine" or "breast_cancer".

• Load the dataset as X and y, and split with train_test_split(X, y, test_size=0.25, random_state=0, stratify=y).
• Build three versions, each ending in KNeighborsClassifier(n_neighbors=k):
  • "none": just the model.
  • "minmax": a pipeline of MinMaxScaler() then the model.
  • "standard": a pipeline of StandardScaler() then the model.
• Fit each on the training rows and score it on the test rows, rounded to 3 decimal places.
• Return a dict with those three scores under the keys "none", "minmax" and "standard", plus "best": the key with the highest rounded score. On a tie, prefer the one that comes first in the order "none", "minmax", "standard".