Source code for xaicompare.adapters.explainers.explainer_shap_tree

# xaicompare/adapters/explainers/explainer_shap_tree.py

import numpy as np
import shap
from scipy.sparse import spmatrix
from typing import Any, Dict, List, Tuple

from xaicompare.adapters.explainers.explainer_base import ExplainerAdapter
from xaicompare.registry.xai_registry import register_xai



@register_xai("shap_tree")
class ShapTreeExplainerAdapter(ExplainerAdapter):
    """
    SHAP Tree Explainer with:
      - Safe sparse handling
      - Forced vectorization
      - Guaranteed 2-D input to SHAP
      - Batch processing to avoid memory errors
      - Multi-class normalization
    """



    def limitation_text(self):
        return "This is some test text about the limitations of the SHAP Tree Explainer."


    def __init__(self, model_adapter, config: Dict[str, Any]):
        super().__init__(model_adapter, config)





    # ----------------------------------------------------------------------
    # GLOBAL IMPORTANCE
    # ----------------------------------------------------------------------



    def global_importance(self, X, rows_limit: int = 200) -> Tuple[np.ndarray, List[str]]:
        """
        Compute mean|SHAP| across first rows_limit samples safely.
        """

        # Extract raw values from numpy/pandas objects



    # ----------------------------------------------------------------------
    # LOCAL EXPLANATIONS
    # ----------------------------------------------------------------------



    def local_explanations(self, x_row) -> np.ndarray:
        """
        Compute SHAP for a single example → return signed vector.
        """


        # Force shape (1, raw_item)