Cross-validation (K-fold) in ML Python - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Why use K-fold cross-validation?

Imagine you have a small dataset and want to estimate how well your machine learning model will perform on new data. Why is K-fold cross-validation a better choice than a single train-test split?

A It uses all data points for both training and testing, reducing bias in performance estimates.

B It trains the model multiple times on the same training set to improve accuracy.

C It splits data randomly once, which is faster and more reliable than multiple splits.

D It only tests the model on the largest portion of data to get the best score.

❓ Predict Output

intermediate

Output of K-fold split indices

What will be the output of the following Python code that uses KFold from scikit-learn?

ML Python

from sklearn.model_selection import KFold
import numpy as np

X = np.array([10, 20, 30, 40, 50])
kf = KFold(n_splits=2, shuffle=False)

splits = []
for train_index, test_index in kf.split(X):
    splits.append((train_index.tolist(), test_index.tolist()))

print(splits)

A [([3, 4], [0, 1, 2]), ([0, 1, 2], [3, 4])]

B [([0, 1, 2], [3, 4]), ([3, 4], [0, 1, 2])]

C [([1, 2, 3], [0, 4]), ([0, 4], [1, 2, 3])]

D [([2, 3, 4], [0, 1]), ([0, 1], [2, 3, 4])]

❓ Model Choice

advanced

Choosing K for K-fold cross-validation

You have a dataset with 1000 samples. You want to use K-fold cross-validation to estimate model performance. Which choice of K balances bias and variance best?

A K = 5, because it provides a good balance between bias and variance for many datasets.

B K = 1000, because leave-one-out cross-validation always gives the best estimate.

C K = 1, because using the whole dataset for training and testing is most accurate.

D K = 2, because fewer folds reduce computation time and variance.

❓ Metrics

advanced

Calculating average accuracy from K-fold results

You performed 4-fold cross-validation and got these accuracy scores for each fold: [0.82, 0.85, 0.80, 0.83]. What is the correct average accuracy to report?

A 0.8250

B 0.83

C 0.825

D 0.83 with standard deviation 0.02

🔧 Debug

expert

Why does this K-fold code raise an error?

Consider this Python code snippet using KFold. It raises an error. What is the cause?

ML Python

from sklearn.model_selection import KFold
import numpy as np

X = np.array([1, 2, 3])
kf = KFold(n_splits=5)

for train_index, test_index in kf.split(X):
    print('Train:', train_index, 'Test:', test_index)

A IndexError because test indices exceed array length.

B ValueError because n_splits (5) cannot be greater than the number of samples (3).

C TypeError because KFold expects a list, not a numpy array.

D RuntimeError because KFold requires shuffle=true for small datasets.