Bias detection and mitigation in ML Python - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to calculate the demographic parity difference.

ML Python

demographic_parity_diff = abs(P(y_pred = 1 | [1] = 0) - P(y_pred = 1 | [1] = 1))

Drag options to blanks, or click blank then click option'

Atarget

Bsensitive_attribute

Cfeature

Dprediction

Attempts:

3 left

💡 Hint

Common Mistakes

Using the target variable instead of the sensitive attribute.

Using the prediction variable instead of the sensitive attribute.

The demographic parity difference measures the difference in positive prediction rates between groups defined by the sensitive attribute.

2fill in blank

medium

Complete the code to split data into training and test sets with stratification on the sensitive attribute.

ML Python

train_test_split(X, y, test_size=0.3, stratify=[1], random_state=42)

Drag options to blanks, or click blank then click option'

Asensitive_attribute

BX

Cy

Drandom_state

Attempts:

3 left

💡 Hint

Common Mistakes

Stratifying on the target variable instead of the sensitive attribute.

Not stratifying at all, causing imbalance.

Stratifying on the sensitive attribute ensures that both training and test sets have similar distributions of that attribute.

3fill in blank

hard

Fix the error in the fairness metric calculation by completing the code.

ML Python

equal_opportunity_diff = abs(TPR_[1] - TPR_{group1})

Drag options to blanks, or click blank then click option'

Agroup0

Bgroup1

Cpositive

Dnegative

Attempts:

3 left

💡 Hint

Common Mistakes

Using the same group for both TPRs.

Using positive or negative instead of group names.

True Positive Rate (TPR) should be compared between two groups, here group0 and group1. The code needs to specify the first group in the first TPR.

4fill in blank

hard

Fill both blanks to create a dictionary comprehension that filters features with importance above a threshold.

ML Python

important_features = {feature: importance for feature, importance in feature_importances.items() if importance [1] [2]

Drag options to blanks, or click blank then click option'

A>

B0.05

C<

D0.1

Attempts:

3 left

💡 Hint

Common Mistakes

Using '<' instead of '>' causing wrong filtering.

Using too high or too low threshold.

This comprehension selects features whose importance is greater than 0.05.

5fill in blank

hard

Fill all three blanks to create a fairness-aware model training pipeline.

ML Python

model = [1](sensitive_features=[2])
model.fit(X_train, y_train)
predictions = model.[3](X_test)

Drag options to blanks, or click blank then click option'

AFairClassifier

Bsensitive_attribute

Cpredict

DRandomForestClassifier

Attempts:

3 left

💡 Hint

Common Mistakes

Using a standard classifier without fairness parameters.

Using fit instead of predict for predictions.

FairClassifier is initialized with sensitive features, then trained and used to predict on test data.