SHAP

SHAP (SHapley Additive exPlanations) is a mainstream machine learning model explanation framework whose theoretical basis derives from Shapley values in game theory.

LIME and SHAP are methods in the field of explainable AI (XAI) that attempt to explain model behavior post hoc.

LIME (Local Interpretable Model-Agnostic Explanations) and SHAP (SHapley Additive exPlanations) are the primary tools for providing post-hoc explanations for complex black-box models.

SHAP基于博弈论中的Shapley值，能够量化每个特征对最终预测结果的贡献度，是目前工业界应用最广泛的解释性工具之一

LIME explains individual predictions by generating a perturbed dataset around the input sample to train a local linear approximation model, while SHAP provides feature attribution based on game-theoretic Shapley values; both are model-agnostic techniques.

In 2017, Lundberg and Lee proposed SHAP, introducing the Shapley value into the field of machine learning interpretability.

Providing interpretability information (such as feature attributions generated by LIME or SHAP) can effectively counter automation bias.

Pixel-level attribution methods such as LIME, SHAP, and Grad-CAM use pixels or regions as the unit of explanation rather than semantic concepts familiar to humans.

SHAP and LIME are both post-hoc explanation frameworks that do not modify the model itself, but instead generate explanations after predictions are made.

Commonly used interpretability mitigation measures for medical AI in the industry include attention visualization (Grad-CAM), SHAP value explanations, and "human-in-the-loop" design.

The research integrated four feature importance methods — SHAP, grouped permutation importance, Boruta, and Recursive Feature Elimination (RFE) — and merged the results through a consensus ranking strategy.

The nine evaluated XAI methods cover gradient-based methods (Vanilla Gradient, Integrated Gradients), perturbation-based methods (LIME, KernelSHAP), and propagation-rule-based methods (LRP).

Mainstream methods such as LIME and SHAP treat maximizing local fidelity as their optimization objective by design.

XGBoost has mature feature importance analysis, SHAP value interpretation, and engineering toolchains, whereas Transformer-based tabular foundation models remain immature in terms of interpretability.

Attention weights are not equivalent to causal explanations, SHAP values are computationally expensive in high-dimensional scenarios, and the reasoning process output in chain-of-thought does not necessarily correspond one-to-one with the model's internal computation.

Academia has developed explanation methods such as SHAP values and LIME, but these tools currently struggle to provide complete explanation paths sufficient to support high-stakes decisions.

Researchers used SHAP (Shapley Additive Explanations) to analyze traditional classifiers in order to identify key features driving prediction decisions.

SHAP values (SHapley Additive exPlanations) can provide individual-level feature contribution explanations and integrate well with tree models.

Mainstream explainable AI techniques include post-hoc explanation methods such as SHAP (game-theory-based Shapley value attribution), LIME (local interpretable model-agnostic explanations), and attention mechanism visualization, as well as inherently interpretable models such as decision trees, logistic regression, and rule lists.

SHAP and LIME are post-hoc explanation techniques for addressing algorithmic interpretability challenges.