Handling categorical variables in ML Python - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Why use one-hot encoding for categorical variables?

Why is one-hot encoding commonly used to handle categorical variables in machine learning?

A It reduces the number of features by combining categories into one number.

B It replaces missing values in categorical data with the most frequent category.

C It assigns a unique integer to each category, preserving their natural order.

D It converts categories into a numeric format without implying any order between them.

❓ Predict Output

intermediate

Output of label encoding on a categorical list

What is the output of the following Python code using sklearn's LabelEncoder?

ML Python

from sklearn.preprocessing import LabelEncoder
le = LabelEncoder()
categories = ['red', 'blue', 'green', 'blue', 'red']
encoded = le.fit_transform(categories)
print(list(encoded))

A [0, 0, 1, 1, 0]

B [1, 2, 0, 2, 1]

C [2, 0, 1, 0, 2]

D [0, 1, 2, 1, 0]

❓ Model Choice

advanced

Best model choice for high-cardinality categorical data

You have a dataset with a categorical feature containing 10,000 unique categories. Which model is best suited to handle this feature without extensive preprocessing?

A Decision Tree without encoding the categorical feature

B Gradient Boosting with target encoding of the categorical feature

C Linear Regression with one-hot encoding of the categorical feature

D K-Nearest Neighbors with label encoding of the categorical feature

❓ Hyperparameter

advanced

Choosing encoding method for tree-based models

Which encoding method is generally preferred for categorical variables when using tree-based models like Random Forest or XGBoost?

A Label encoding to assign integers to categories

B Binary encoding to reduce dimensionality

C Frequency encoding to replace categories with their counts

D One-hot encoding to create binary features for each category

❓ Metrics

expert

Evaluating impact of encoding on model performance

You trained two models on the same dataset: Model A uses one-hot encoding for categorical variables, Model B uses target encoding. Both models are gradient boosting classifiers. Model A has 85% accuracy, Model B has 88% accuracy on the test set. What is the most likely explanation?

A Target encoding helped Model B capture category information better, improving accuracy.

B One-hot encoding always causes overfitting, so Model A performs worse.

C Model B's higher accuracy is due to random chance and not encoding choice.

D One-hot encoding is better for gradient boosting, so Model A should have higher accuracy.