What is Explainable AI (XAI)? 30-Second Breakdown

import shap
import pandas as pd

# Load your model and data
model = xgboost.XGBClassifier()
model.fit(X_train, y_train)

# Create SHAP explainer
explainer = shap.TreeExplainer(model)
shap_values = explainer.shap_values(X_test)

# Show feature importance for one prediction
shap.waterfall_plot(explainer.expected_value[1], shap_values[1][0], X_test.iloc[0])

SHAP reveals which features contributed most to each prediction

from lime import lime_text
from sklearn.pipeline import make_pipeline

# Text classification example
explainer = lime_text.LimeTextExplainer(class_names=['negative', 'positive'])

# Explain a single prediction
exp = explainer.explain_instance(
    text_instance, 
    classifier_fn, 
    num_features=10
)
exp.show_in_notebook(text=True)

LIME explains individual predictions by approximating the model locally

def explain_prediction(model, instance, explainer_type='shap'):
    """
    Explain a single model prediction
    """
    if explainer_type == 'shap':
        explainer = shap.Explainer(model)
        shap_values = explainer(instance)
        return shap.plots.waterfall(shap_values[0])
    
    elif explainer_type == 'lime':
        explainer = lime.lime_tabular.LimeTabularExplainer(
            X_train.values,
            feature_names=X_train.columns,
            class_names=['No', 'Yes']
        )
        exp = explainer.explain_instance(instance, model.predict_proba)
        return exp.show_in_notebook()

# Usage
explanation = explain_prediction(my_model, test_instance, 'shap')