Explainable Machine learning Exercises

Dataset

• Bike dataset to predict the number of bikes in a particular day (regression problem)
• Cervical Cancer Dataset to predict Biopsy (classification problem)

Methods and Exploration, What every exercise is about?

CE1-

• Random Forest regressor, Adjusted R2 Score
• Generation of effect plot of the Lasso algorithm
• Interaction feature and Linear Regression assumptions

CE2-

• Logistic Regression, Odds ratio
• Decision Trees, Feature importance from Mean Decrease in Impurity
• RuleFit - Extract rules using CART and adding those rules as features then using LASSO
• Extract highest support attribute from rules

CE3-

• Partial Dependence Plots
• Individual Conditional Expectation plots
• M-plot
• Accumulated local effects
• Permutation feature importance

CE4 -

• Local post-hoc explanations
  1. Counterfactual explanations
  2. Local interpretable model-agnostic explanations
  3. Shapley Additive explanations, Kernel SHAP

[2023]