Stratified K-fold in ML Python - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Why use Stratified K-fold instead of regular K-fold?

Imagine you have a dataset with two classes: 90% are class A and 10% are class B. You want to split the data into 5 folds for cross-validation.

Why is Stratified K-fold better than regular K-fold in this case?

A It ensures each fold has approximately the same percentage of samples of each class as the whole dataset.

B It randomly shuffles data without considering class distribution, which is faster.

C It creates folds with only one class each to simplify training.

D It duplicates minority class samples to balance the dataset before splitting.

❓ Predict Output

intermediate

Output of StratifiedKFold split indices

Given this code, what is the output of the printed train indices for the first fold?

ML Python

from sklearn.model_selection import StratifiedKFold
import numpy as np

X = np.array([[i] for i in range(10)])
y = np.array([0,0,0,0,1,1,1,1,1,1])

skf = StratifiedKFold(n_splits=2, shuffle=False)

for fold, (train_index, test_index) in enumerate(skf.split(X, y)):
    if fold == 0:
        print(train_index.tolist())

A [2, 3, 7, 8, 9]

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

C [5, 6, 7, 8, 9]

D [0, 1, 2, 3, 4, 5]

❓ Model Choice

advanced

Choosing the best cross-validation method for imbalanced data

You have a dataset with 95% of samples in class 0 and 5% in class 1. You want to evaluate a classification model's performance reliably.

Which cross-validation method is best to use?

A Leave-One-Out cross-validation

B Random train-test split without cross-validation

C Stratified K-fold cross-validation

D Regular K-fold cross-validation without stratification

❓ Hyperparameter

advanced

Effect of increasing n_splits in StratifiedKFold

What is the effect of increasing the number of splits (n_splits) in StratifiedKFold on the training and validation sets?

A Both training and validation sets remain the same size regardless of n_splits.

B Training sets become smaller and validation sets become smaller, leading to overfitting.

C Training sets become smaller and validation sets become larger, increasing variance in performance estimates.

D Training sets become larger and validation sets become smaller, reducing bias in performance estimates.

🔧 Debug

expert

Why does this StratifiedKFold code raise an error?

Consider this code snippet:

from sklearn.model_selection import StratifiedKFold
import numpy as np

X = np.array([[i] for i in range(6)])
y = np.array([0, 0, 1, 1, 1, 1])

skf = StratifiedKFold(n_splits=3)

for train_index, test_index in skf.split(X, y):
    print("TRAIN:", train_index, "TEST:", test_index)

Running this code raises a ValueError. What is the cause?

A The input arrays X and y have mismatched lengths.

B The number of splits is greater than the number of members in the smallest class.

C StratifiedKFold requires shuffle=True when n_splits > 2.

D The labels y contain non-integer values which StratifiedKFold cannot handle.