Fruit are described by weight (hundreds of grams) and sweetness (a small score). Without scaling, weight drowns out sweetness. See whether scaling changes the answers.
Task: write raw_and_scaled(train_rows, train_labels, new_rows). Each row is [weight, sweetness].
KNeighborsClassifier(n_neighbors=1) on the rows as they are, and predict new_rows.MinMaxScaler on train_rows only, transform both train_rows and new_rows with it, train a fresh KNeighborsClassifier(n_neighbors=1) on the scaled training rows, and predict the scaled new rows.[raw predictions, scaled predictions], each a list in the order of new_rows.