Model Interpretability, Titanic

The workflow demonstrates how to use SHAP, Shapley Values and LIME implemenatations in KNIME 4.0 and generates a basic combined view.

It trains a Random Forest model for predicting survival of the Titanic dataset and compute explanations using those three different techniques.

The general steps demonstrated in this workflow are to:

1) Clean the data
2) Train the model
3) Take a sample row to explain
4) Run SHAP, Shapley Values and LIME
5) Combine the results in an interactive composite view.

External resources

• Christoph Molnar - Interpretable Machine Learning - A Guide for Making Black Box Models Explainable - 5.10 SHAP (SHapley Additive exPlanations)
• Christoph Molnar - Interpretable Machine Learning - A Guide for Making Black Box Models Explainable - 5.9 Shapley Values
• Christoph Molnar - Interpretable Machine Learning - A Guide for Making Black Box Models Explainable - 5.7 Local Surrogate (LIME)