nananie143's picture
feat: Complete blueprint implementation with 66+ modules
90bacf7 verified
"""
LIME Explainer Module
Local interpretable model explanations.
Part of the complete blueprint implementation.
"""
import numpy as np
from typing import Dict, List, Optional, Any
import logging
logger = logging.getLogger(__name__)
try:
import lime
from lime.lime_tabular import LimeTabularExplainer
LIME_AVAILABLE = True
except ImportError:
LIME_AVAILABLE = False
class LIMEExplainer:
"""
LIME-based model explanation.
Features:
- Local explanations
- Model-agnostic
- Feature contributions
"""
def __init__(self):
self.explainer = None
self.feature_names = []
self.class_names = ['H', 'D', 'A']
def fit(
self,
X_train: np.ndarray,
feature_names: List[str] = None,
class_names: List[str] = None,
mode: str = 'classification'
) -> 'LIMEExplainer':
"""
Fit LIME explainer.
"""
self.feature_names = feature_names or [f'feature_{i}' for i in range(X_train.shape[1])]
self.class_names = class_names or self.class_names
self.mode = mode
if not LIME_AVAILABLE:
logger.warning("LIME not available")
return self
try:
self.explainer = LimeTabularExplainer(
X_train,
feature_names=self.feature_names,
class_names=self.class_names,
mode=mode
)
logger.info("LIME explainer fitted successfully")
except Exception as e:
logger.warning(f"Could not create LIME explainer: {e}")
return self
def explain_prediction(
self,
X: np.ndarray,
predict_fn: callable,
num_features: int = 10
) -> Dict:
"""
Explain a single prediction.
"""
if self.explainer is None or not LIME_AVAILABLE:
return self._fallback_explanation()
try:
if len(X.shape) > 1:
X = X[0]
explanation = self.explainer.explain_instance(
X,
predict_fn,
num_features=num_features
)
# Extract features
feature_weights = explanation.as_list()
return {
'top_features': [
{'feature': f, 'contribution': round(w, 4)}
for f, w in feature_weights
],
'positive_contributors': [
{'feature': f, 'contribution': round(w, 4)}
for f, w in feature_weights if w > 0
],
'negative_contributors': [
{'feature': f, 'contribution': round(w, 4)}
for f, w in feature_weights if w < 0
],
'local_prediction': explanation.local_pred[0] if hasattr(explanation, 'local_pred') else None
}
except Exception as e:
logger.warning(f"LIME explanation failed: {e}")
return self._fallback_explanation()
def _fallback_explanation(self) -> Dict:
"""Fallback when LIME unavailable."""
return {
'method': 'unavailable',
'top_features': [],
'message': 'LIME not available or explanation failed'
}
def generate_html_explanation(
self,
X: np.ndarray,
predict_fn: callable
) -> str:
"""Generate HTML explanation."""
if self.explainer is None or not LIME_AVAILABLE:
return "<p>LIME not available</p>"
try:
explanation = self.explainer.explain_instance(X[0] if len(X.shape) > 1 else X, predict_fn)
return explanation.as_html()
except Exception:
return "<p>Explanation failed</p>"
_explainer: Optional[LIMEExplainer] = None
def get_explainer() -> LIMEExplainer:
global _explainer
if _explainer is None:
_explainer = LIMEExplainer()
return _explainer