Model interpretability (SHAP, LIME) in ML Python - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to import the SHAP library.

ML Python

import [1]

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

Ashap

Blime

Csklearn

Dpandas

Attempts:

3 left

💡 Hint

Common Mistakes

Trying to import 'lime' instead of 'shap'.

Importing unrelated libraries like 'pandas' or 'sklearn'.

The SHAP library is imported using import shap. This library helps explain model predictions.

2fill in blank

medium

Complete the code to create a LIME explainer for tabular data.

ML Python

from lime import lime_tabular
explainer = lime_tabular.LimeTabularExplainer([1], feature_names=feature_names, class_names=class_names, mode='classification')

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

Ay_train

BX_train

Cmodel

DX_test

Attempts:

3 left

💡 Hint

Common Mistakes

Passing labels (y_train) instead of features.

Passing the model object instead of data.

The LIME explainer needs the training data features (X_train) to learn the data distribution for explanations.

3fill in blank

hard

Fix the error in the SHAP values calculation code.

ML Python

explainer = shap.TreeExplainer(model)
shap_values = explainer.[1](X_test)

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

Ashap_values

Bexplain

Ccalculate

Dtransform

Attempts:

3 left

💡 Hint

Common Mistakes

Using non-existent methods like 'transform' or 'calculate'.

Trying to call 'explain' which is not a method here.

The correct method to get SHAP values from the explainer is shap_values().

4fill in blank

hard

Fill both blanks to create a dictionary of feature importances from SHAP values.

ML Python

feature_importance = {feature: abs(shap_values[[1]][i]) for i, feature in enumerate([2])}

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

A0

B1

Cfeature_names

DX_test

Attempts:

3 left

💡 Hint

Common Mistakes

Using index 1 which may correspond to another class.

Using raw data X_test instead of feature names.

SHAP values for classification models are often in a list where index 0 corresponds to the output class. Feature names are used to map values.

5fill in blank

hard

Fill all three blanks to generate a LIME explanation and show the predicted class.

ML Python

exp = explainer.explain_instance([1], model.predict_proba, num_features=[2])
predicted_class = model.predict([[3]])[0]

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

AX_test[0]

B5

DX_train[0]

Attempts:

3 left

💡 Hint

Common Mistakes

Using training data instead of test data for explanation.

Passing wrong number of features to show.

The LIME explainer needs a single instance from test data (X_test[0]), number of features to show (5), and the model prediction on that instance.