Attribute model predictions to input features using SHAP values.
npx skills add https://github.com/crazymsn/academic-skills --skill shap-crazymsn
Please help me install this Agent Skill. Skill: shap Source: https://github.com/crazymsn/academic-skills/tree/main/academic-skills/shap Command: npx skills add https://github.com/crazymsn/academic-skills --skill shap-crazymsn
SHAP provides a principled framework to explain model predictions by attributing outputs to individual input features using Shapley values, enabling transparent decision-making and debugging.
Run an end-to-end explanation by loading a trained model, selecting a suitable explainer, computing SHAP values for a sample, and visualizing global feature importances with a beeswarm plot.
High-intent search queries and answers about installing and using this skill.
Feature attribution explains model predictions by assigning output contributions to input features using Shapley values. This framework enables transparent decision-making, model debugging, and fairness analysis across various black-box models.
TreeExplainer is designed for tree-based models, while DeepExplainer and GradientExplainer target deep learning models. KernelExplainer and PermutationExplainer apply to black-box models, and LinearExplainer handles linear models.
Model feature importances can be visualized using beeswarm plots for global interpretation, waterfall plots for individual predictions, scatter plots, and heatmaps to understand feature attribution distributions across samples.
A defined background dataset is required to compute SHAP values for model interpretation. It serves as the reference distribution for attributing output changes to input features across the selected explainer.
SHAP values support fairness analysis and model comparison by quantifying feature attribution across different models. They reveal how input features influence predictions, helping identify bias and compare model behaviors.
Yezez9Compute SHAP feature attributions and generate explanation plots for machine learning models.
xiechyCompute SHAP values to attribute predictions to features across model types.
shushuznCompute SHAP values and generate interpretability visualizations for machine learning models.
pur3v4d3rAttribute model predictions to input features using SHAP explainers.
