Feature engineering pipelines in MLOps - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

What is the main purpose of a feature engineering pipeline in MLOps?

Choose the best description of why we use feature engineering pipelines in machine learning operations.

A To collect raw data from various sources.

B To deploy machine learning models to production environments.

C To automate and standardize the process of transforming raw data into features for models.

D To monitor the performance of models after deployment.

💡 Hint

Think about what happens before training a model with raw data.

Feature engineering pipelines automate and standardize how raw data is transformed into features, ensuring consistency and reproducibility.

💻 Command Output

intermediate

Output of a feature pipeline step using scikit-learn's ColumnTransformer

Given the following Python code snippet, what is the output shape of X_transformed?

MLOps

from sklearn.compose import ColumnTransformer
from sklearn.preprocessing import StandardScaler, OneHotEncoder
import numpy as np

X = np.array([[25, 'red'], [30, 'blue'], [22, 'green']])

preprocessor = ColumnTransformer(
    transformers=[
        ('num', StandardScaler(), [0]),
        ('cat', OneHotEncoder(), [1])
    ])

X_transformed = preprocessor.fit_transform(X)
print(X_transformed.shape)

A (3, 3)

B (3, 4)

C (2, 4)

D (3, 2)

💡 Hint

Count numeric and categorical features after transformation.

The numeric column is scaled to 1 feature, and the categorical column with 3 unique values is one-hot encoded to 3 features, total 4 features.

🔀 Workflow

advanced

Order the steps to build a feature engineering pipeline for a new dataset

Arrange the following steps in the correct order to create a feature engineering pipeline.

A 3,1,2,4

B 2,1,3,4

C 1,3,2,4

D 1,2,3,4

💡 Hint

Think about understanding data first, then defining transformations, then implementation, then testing.

First identify data and types, then define transformations, implement the pipeline, and finally test it.

❓ Troubleshoot

advanced

Why does this feature pipeline raise a ValueError during fit?

Consider this code snippet that raises an error during fit. What is the most likely cause?

MLOps

from sklearn.pipeline import Pipeline
from sklearn.preprocessing import StandardScaler
from sklearn.impute import SimpleImputer
import numpy as np

X = np.array([[1, 2], [np.nan, 3], [7, 6]])

pipeline = Pipeline([
    ('imputer', SimpleImputer(strategy='mean')),
    ('scaler', StandardScaler())
])

pipeline.fit(X)

A Pipeline steps are in wrong order; imputer should come before scaler.

B SimpleImputer requires categorical data, but numeric data was given.

C StandardScaler cannot handle NaN values before imputation.

D The input array X has inconsistent row lengths.

💡 Hint

Think about which step should handle missing values first.

Imputation must happen before scaling because StandardScaler cannot process NaN values.

✅ Best Practice

expert

Which practice ensures feature engineering pipelines support model reproducibility?

Choose the best practice that helps maintain reproducibility of machine learning models when using feature engineering pipelines.

A Version control the pipeline code and store pipeline artifacts with the model.

B Run the pipeline only on training data and ignore test data transformations.

C Manually apply transformations outside the pipeline for flexibility.

D Use random transformations without fixing seeds to increase data variety.

💡 Hint

Think about how to keep track of changes and ensure the same transformations are applied later.

Version controlling pipeline code and storing artifacts ensures the exact transformations can be reproduced for future use.