How to Use SHAP and LIME for Model Interpretability

 1. Introduction to SHAP and LIME

SHAP explains model predictions using game theory (Shapley values). It attributes the contribution of each feature to a prediction in a consistent and theoretically sound way.

LIME explains a prediction by approximating the model locally with an interpretable (usually linear) model.

2. When to Use SHAP and LIME

Criteria SHAP LIME

Theoretical soundness ✅ Strong theoretical foundation ❌ Heuristic-based

Speed ❌ Can be slower ✅ Faster (approximate)

Local explanations ✅ Yes ✅ Yes

Global interpretability ✅ Can be aggregated ❌ Primarily local

Model agnostic ✅ (via KernelExplainer) ✅ Yes

3. How to Use SHAP

Installation

bash

Copy

Edit

pip install shap

Basic Usage (for tree-based models like XGBoost, LightGBM)

python

Copy

Edit

import shap

import xgboost

import pandas as pd

# Train a model

X, y = shap.datasets.boston()

model = xgboost.XGBRegressor().fit(X, y)

# Create SHAP explainer

explainer = shap.Explainer(model)

# Compute SHAP values

shap_values = explainer(X)

# Visualize SHAP values for a single prediction

shap.plots.waterfall(shap_values[0])

# Visualize summary

shap.plots.beeswarm(shap_values)

4. How to Use LIME

Installation

bash

Copy

Edit

pip install lime

Basic Usage

python

Copy

Edit

import lime

import lime.lime_tabular

from sklearn.ensemble import RandomForestClassifier

from sklearn.datasets import load_iris

# Load data and train model

iris = load_iris()

X, y = iris.data, iris.target

model = RandomForestClassifier().fit(X, y)

# Create LIME explainer

explainer = lime.lime_tabular.LimeTabularExplainer(

    X_train=X,

    feature_names=iris.feature_names,

    class_names=iris.target_names,

    mode='classification'

)

# Explain a prediction

i = 0

exp = explainer.explain_instance(X[i], model.predict_proba, num_features=4)

# Show explanation

exp.show_in_notebook()

5. Visualizations

SHAP:

waterfall plot: Shows feature contribution for a single prediction.

beeswarm plot: Shows distribution of SHAP values for all predictions.

bar plot: Shows mean absolute SHAP values per feature.

LIME:

Interactive HTML showing local decision boundaries and feature weights.

6. Tips and Best Practices

Use SHAP for consistency and accuracy, especially with tree-based models.

Use LIME when speed is more important, or for quick prototyping.

Combine with domain knowledge for better trust in the model.

Always use explanations to validate and debug models, not as a substitute for understanding the data or the model itself.

7. Conclusion

Both SHAP and LIME help make black-box models more interpretable by explaining why a prediction was made. Use them to:

Build trust with stakeholders.

Debug model decisions.

Identify biased or unexpected behavior.

Learn Data Science Course in Hyderabad

Read More

Outlier Detection Methods in Data Science

How to Handle Categorical Data in Machine Learning Models

Feature Selection Techniques: Filter, Wrapper, and Embedded Methods

How to Use Principal Component Analysis (PCA) for Dimensionality Reduction

Visit Our Quality Thought Training Institute in Hyderabad

Get Directions