Example Explanations
Below are representative examples of the visual explanations for a single prediction from the Titanic dataset offered by the three included explainers: LIME, SHAP, and Anchor.
LIME
Prediction Probabilities: Likelihood of each possible classification output returned by the model
Local Feature Weights: Numbers below features are weights representing how much each feature contributed to the "Yes" or "No" class (survived or did not survived).
Shap
The size of the arrows represent the magnitude of the contribution of features. Blue colors represent negative contribution while red colors represent positive contribution of this record being classified as "Yes" / "Survived".
Anchor
"Sex = male" is the anchor of the explanation and has 84% precision for the perturbation space