Package artemis
artemis
: A Robust Toolkit of Explanation Methods for Interaction Spotting
What is artemis
?
artemis
is a Python library for explanations of feature interactions in machine learning models.
- It provides methods for analyzing predictive models in terms of interactions between features and feature importance.
- There are both model-agnostic methods that can work with any properly prepared model, and model-specific methods adapted to tree-based ones
(due to their structure, which naturally influences the possibility of interaction).
- It enables to scrutinize a wide range of models by examining strength of feature interactions and visualizing them.
Expand source code
"""
# `artemis`: A Robust Toolkit of Explanation Methods for Interaction Spotting
## What is `artemis`?
**`artemis` is a Python library for explanations of feature interactions in machine learning models.**
- It provides methods for analyzing predictive models in terms of interactions between features and feature importance.
- There are both model-agnostic methods that can work with any properly prepared model, and model-specific methods adapted to tree-based ones
(due to their structure, which naturally influences the possibility of interaction).
- It enables to scrutinize a wide range of models by examining strength of feature interactions and visualizing them.
"""
Sub-modules
artemis.additivity
artemis.comparison
artemis.importance_methods
artemis.interactions_methods
artemis.visualizer