Lime | Model Interpretability | Machine Learning Tutorial

Introduction

Model interpretability is a crucial aspect of machine learning, especially when the models are used in critical applications such as healthcare, finance, and law. LIME (Local Interpretable Model-Agnostic Explanations) is a technique used to interpret the predictions of any machine learning model by approximating it locally with an interpretable model.

What is LIME?

LIME is a tool that explains the predictions of black-box models by approximating them with simple, interpretable models. It provides local explanations, meaning it explains the predictions for individual instances rather than the model as a whole. This allows users to understand why a model made a particular prediction for a specific instance.

How LIME Works

LIME works by generating a new dataset consisting of perturbed samples around the instance of interest and obtaining predictions for these samples from the black-box model. It then trains an interpretable model (such as a linear model) on this new dataset to approximate the predictions of the black-box model locally. The interpretable model is then used to explain the prediction for the instance of interest.

Installing LIME

To use LIME, you need to install the lime package. You can install it using pip:

pip install lime

Using LIME

Let's go through an example to understand how to use LIME for explaining the predictions of a machine learning model.

Example: Explaining a Classifier

We will use the Iris dataset and a RandomForestClassifier to demonstrate how LIME can be used to explain the predictions of the classifier.

First, let's import the necessary libraries and load the dataset:

import numpy as np
import pandas as pd
from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestClassifier
from lime.lime_tabular import LimeTabularExplainer

# Load the Iris dataset
iris = load_iris()
X = iris.data
y = iris.target

# Split the dataset into training and testing sets
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

Next, let's train a RandomForestClassifier on the training data:

# Train a RandomForestClassifier
clf = RandomForestClassifier(n_estimators=100, random_state=42)
clf.fit(X_train, y_train)

Now, let's use LIME to explain the predictions of our classifier. We will create a LimeTabularExplainer and use it to explain a prediction for a single instance from the test set:

# Create a LimeTabularExplainer
explainer = LimeTabularExplainer(X_train, feature_names=iris.feature_names, class_names=iris.target_names, discretize_continuous=True)

# Choose an instance to explain
instance = X_test[0]

# Generate an explanation for the instance
explanation = explainer.explain_instance(instance, clf.predict_proba, num_features=2)

# Show the explanation
explanation.show_in_notebook(show_table=True, show_all=False)

The output will be a human-readable explanation of the model's prediction for the chosen instance, showing the contribution of each feature to the prediction.

Feature Contributions:

petal width (cm): 0.45
petal length (cm): 0.30

Conclusion

LIME is a powerful tool for understanding and interpreting the predictions of black-box models. By approximating the model locally with an interpretable model, LIME provides insights into why a model made a particular prediction for a specific instance. This can help build trust in the model and identify potential issues.

LIME: Local Interpretable Model-Agnostic Explanations