Fill, Scale, Vote

A clinic's records have gaps. A three-step pipeline can fill them, scale the columns and classify, all from one fit.

Task: write predict_with_gaps(train_rows, train_labels, new_rows).

• Rows are lists of numbers, with None where a value is missing. Turn them into arrays with np.array(rows, dtype=float), which makes each None a np.nan.
• Build a Pipeline with three steps, in this order:
  1. ("fill", SimpleImputer(strategy="median"))
  2. ("scale", StandardScaler())
  3. ("model", KNeighborsClassifier(n_neighbors=3))
• Fit it on the training rows, then predict new_rows. New rows can have gaps too.
• Return the predictions as a list.