Creating interaction features in ML Python - Practice Exercises

Challenge - 5 Problems

🧠 Conceptual

intermediate

Why create interaction features in a model?

Imagine you have two features: hours studied and hours slept. Why might creating an interaction feature (like multiplying these two) help your model?

A Because the combined effect of studying and sleeping might influence the outcome differently than each alone

B Because interaction features always reduce the number of features and simplify the model

C Because multiplying features always fixes missing data problems

D Because interaction features remove the need for normalization

💡 Hint

Think about how two things working together might have a different effect than each separately.

Interaction features capture how two variables together affect the target differently than individually. For example, studying a lot but sleeping very little might have a different impact than studying and sleeping both moderately.

❓ Predict Output

intermediate

Output of interaction feature creation code

What is the output of this Python code that creates an interaction feature?

ML Python

import pandas as pd

df = pd.DataFrame({'A': [1, 2, 3], 'B': [4, 5, 6]})
df['A_B'] = df['A'] * df['B']
print(df['A_B'].tolist())

A [1, 2, 3]

B [5, 7, 9]

C [4, 10, 18]

D [4, 5, 6]

💡 Hint

Multiply each element of column A by the corresponding element in column B.

The new column 'A_B' is created by multiplying values in 'A' and 'B' row-wise: 1*4=4, 2*5=10, 3*6=18.

❓ Model Choice

advanced

Best model type for interaction features

You have created many interaction features from your dataset. Which model type is best suited to automatically capture complex interactions without explicitly creating interaction features?

A K-Nearest Neighbors without feature scaling

B Linear Regression without interaction terms

C Simple Logistic Regression without polynomial features

D Decision Trees or Random Forests

💡 Hint

Think about models that split data based on feature values and can capture combinations naturally.

Decision Trees and Random Forests can capture complex feature interactions by splitting data on multiple features in sequence, without needing explicit interaction features.

❓ Hyperparameter

advanced

Hyperparameter affecting interaction feature complexity

When using polynomial features to create interaction terms, which hyperparameter controls the maximum degree of interactions included?

A degree

B alpha

C max_depth

D learning_rate

💡 Hint

It controls how many features are multiplied together.

The degree parameter in polynomial feature generation controls the highest power of features and thus the maximum interaction order included.

❓ Metrics

expert

Effect of interaction features on model metrics

You add interaction features to a regression model. After training, the training error decreases but the validation error increases. What does this indicate?

A The model is underfitting and needs more features

B The model is overfitting due to too many interaction features

C The interaction features improved generalization

D The data has no noise and the model is perfect

💡 Hint

Think about what it means when training error goes down but validation error goes up.

When training error decreases but validation error increases, the model fits training data too closely and fails to generalize, indicating overfitting often caused by too many or complex features like interaction terms.