user@devops:~$ cat README.md
XAI: SHAP & LIME
# Description
XAI: SHAP & LIME is a complete Explainable AI project demonstrating how to interpret Machine Learning models. Covers SHAP (SHapley Additive exPlanations) with TreeExplainer for Random Forest and XGBoost, generating beeswarm summary plots, global importance bar plots, dependence plots for the top 4 features, and waterfall plots decomposing individual predictions. Includes LIME (Local Interpretable Model-agnostic Explanations) with local explanations showing how each value range contributes to specific predictions. Compares 3 importance methods: Random Forest built-in, scikit-learn Permutation Importance, and SHAP mean |SHAP|. For classification, applies multiclass SHAP to the Wine dataset to visualize which features discriminate each varietal. 14 visualizations generated with matplotlib and seaborn.
# Key features
$ SHAP TreeExplainer for Random Forest and XGBoost with 200 explanation samples
$ SHAP Summary Beeswarm: distribution of each feature's impact on predictions
$ SHAP Dependence Plots: feature → SHAP value relationship for top 4 variables
$ SHAP Waterfall: visual decomposition of individual predictions into contributions
$ LIME Tabular Explainer: local explanations with weighted surrogate models
$ Comparison of 3 importance methods: RF built-in vs Permutation vs SHAP
$ Multiclass SHAP for wine classification with XGBoost (3 classes, 100% accuracy)
$ Analysis of geographic features as price drivers in California housing
# Gallery
# Technologies used
