LIME explains individual predictions by approximating the model locally with an interpretable model and calculating the feature importance based on the surrogate model.
Key Points
-
Purpose: LIME focuses on explaining individual predictions by approximating the model locally using a simpler, interpretable model (like linear regression).
-
How it Works:
- For a given prediction, LIME generates perturbed samples (e.g., by modifying input features).
- It observes how the predictions change, thus inferring feature importance for that specific instance.
-
Use Cases: Useful for understanding why a specific decision was made in complex black-box models.
-
Advantage:
- LIME can work with any model type.
- It is relatively easy to apply to tabular, text, and image data.
-
Scenario:
- A healthcare provider uses a deep learning model to classify whether patients have a high or low risk of heart disease based on several health metrics, such as cholesterol levels, age, and blood pressure.
-
- LIME Explanation: LIME is used to explain why the model flagged a specific patient as high-risk. By perturbing the input data (e.g., altering cholesterol levels and re-running the prediction), LIME shows that the patient’s high cholesterol level and advanced age are the main reasons for the high-risk classification. This makes it easier for the healthcare provider to justify the decision to recommend lifestyle changes or further medical tests.
Example Code
To use LIME for feature importance, you can use the LIME package:
import lime
import lime.lime_tabular
# Create a LIME explainer
explainer = lime.lime_tabular.LimeTabularExplainer(X_train, feature_names=feature_names)
# Explain a single prediction
explanation = explainer.explain_instance(X_test[0], model.predict_proba)
explanation.show_in_notebook()