AI 용어집
인공지능 완전 사전
Feature Interaction Importance
A measure that quantifies the importance of interactions between features in the model's predictions, beyond their individual effects.
Global Feature Importance
A metric that evaluates the importance of each feature across all model predictions, revealing the most influential factors globally.
Model-Agnostic Interpretation
Interpretation techniques that work on any type of model without requiring access to its internal structure, offering maximum flexibility.
Global SHAP Values
The aggregation of SHAP values across the entire dataset to understand the global impact of each feature on the model's predictions.
Conditional Dependence Plot
A visualization that shows the relationship between a feature and the model's predictions while conditioning on the values of other features.
Model-Agnostic Counterfactual Explanations
A method that generates hypothetical examples to explain how the model's predictions would change if the input features were modified.
Model-Agnostic Rule Extraction
A technique that extracts interpretable rules from the global behavior of a black box model without requiring access to its internal structure.
Global Feature Effect
The analysis of the global effect of a feature on the model's predictions, taking into account all its interactions with other features.
Model-Agnostic Feature Importance
A method that evaluates feature importance without depending on the model's structure, using techniques like permutation or elimination.
Global Model Interpretation
The analysis of a model's global behavior to understand how it makes decisions on average across all data.
Model-Agnostic Partial Dependence
A technique that calculates the partial dependence of the model's predictions with respect to a feature, without depending on the model's structure.
Global Feature Interaction
The analysis of interactions between features on a global scale to understand how they collectively influence the model's predictions.
Model-Agnostic Feature Effect
A method that evaluates the effect of a feature on the model's predictions without depending on its internal structure.
Global Model Explanation
A complete explanation of a model's global behavior, including feature importance, their effects, and interactions.