Spaces:
Sleeping
Sleeping
File size: 14,471 Bytes
e478478 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 | """
Explainability Module - InsightGenAI
====================================
SHAP-based model explainability with feature importance plots,
summary plots, and individual prediction explanations.
Author: InsightGenAI Team
Version: 1.0.0
"""
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from typing import Dict, List, Tuple, Optional, Any, Union
import streamlit as st
import warnings
warnings.filterwarnings('ignore')
# Try to import shap, handle if not available
try:
import shap
SHAP_AVAILABLE = True
except ImportError:
SHAP_AVAILABLE = False
class ExplainabilityEngine:
"""
Model explainability engine using SHAP values.
Attributes:
model: Trained model to explain
X: Feature matrix
explainer: SHAP explainer object
shap_values: Calculated SHAP values
"""
def __init__(self, model, X: pd.DataFrame, feature_names: Optional[List[str]] = None):
"""
Initialize the Explainability Engine.
Args:
model: Trained model
X: Feature data (sample for background)
feature_names: List of feature names
"""
if not SHAP_AVAILABLE:
raise ImportError("SHAP is not installed. Please install with: pip install shap")
self.model = model
self.X = X.copy() if isinstance(X, pd.DataFrame) else pd.DataFrame(X)
self.feature_names = feature_names or self.X.columns.tolist()
self.X.columns = self.feature_names
self.explainer = None
self.shap_values = None
self.expected_value = None
# Initialize SHAP explainer
self._init_explainer()
def _init_explainer(self) -> None:
"""Initialize the appropriate SHAP explainer for the model."""
try:
# Try TreeExplainer first (for tree-based models)
self.explainer = shap.TreeExplainer(self.model)
self.shap_values = self.explainer.shap_values(self.X)
self.expected_value = self.explainer.expected_value
except Exception:
try:
# Fall back to KernelExplainer
self.explainer = shap.KernelExplainer(self.model.predict, shap.sample(self.X, 100))
self.shap_values = self.explainer.shap_values(self.X)
self.expected_value = self.explainer.expected_value
except Exception as e:
raise RuntimeError(f"Could not initialize SHAP explainer: {str(e)}")
def get_feature_importance(self) -> pd.DataFrame:
"""
Get global feature importance based on mean absolute SHAP values.
Returns:
pd.DataFrame with feature importance
"""
if self.shap_values is None:
raise ValueError("SHAP values not calculated. Please initialize explainer first.")
# Handle different shap_values formats
if isinstance(self.shap_values, list):
# For multi-class, use the mean across all classes
shap_array = np.abs(np.array(self.shap_values)).mean(axis=0).mean(axis=0)
else:
shap_array = np.abs(self.shap_values).mean(axis=0)
importance_df = pd.DataFrame({
'feature': self.feature_names,
'importance': shap_array
}).sort_values('importance', ascending=False)
return importance_df
def plot_summary(self, max_display: int = 15, figsize: Tuple[int, int] = (10, 8)) -> plt.Figure:
"""
Create SHAP summary plot (beeswarm plot).
Args:
max_display: Maximum number of features to display
figsize: Figure size tuple
Returns:
matplotlib Figure object
"""
fig, ax = plt.subplots(figsize=figsize)
# Handle different shap_values formats
if isinstance(self.shap_values, list):
# For multi-class classification, use the first class
shap_values_plot = self.shap_values[0]
else:
shap_values_plot = self.shap_values
shap.summary_plot(
shap_values_plot,
self.X,
feature_names=self.feature_names,
max_display=max_display,
show=False
)
plt.title('SHAP Summary Plot', fontsize=14, fontweight='bold', pad=20)
plt.tight_layout()
return fig
def plot_feature_importance(self, max_display: int = 15,
figsize: Tuple[int, int] = (10, 8)) -> plt.Figure:
"""
Create bar plot of feature importance.
Args:
max_display: Maximum number of features to display
figsize: Figure size tuple
Returns:
matplotlib Figure object
"""
fig, ax = plt.subplots(figsize=figsize)
# Handle different shap_values formats
if isinstance(self.shap_values, list):
shap_values_plot = self.shap_values[0]
else:
shap_values_plot = self.shap_values
shap.summary_plot(
shap_values_plot,
self.X,
feature_names=self.feature_names,
max_display=max_display,
plot_type='bar',
show=False
)
plt.title('SHAP Feature Importance', fontsize=14, fontweight='bold', pad=20)
plt.tight_layout()
return fig
def plot_waterfall(self, instance_idx: int = 0,
max_display: int = 10,
figsize: Tuple[int, int] = (12, 6)) -> plt.Figure:
"""
Create waterfall plot for a single prediction.
Args:
instance_idx: Index of the instance to explain
max_display: Maximum number of features to display
figsize: Figure size tuple
Returns:
matplotlib Figure object
"""
fig, ax = plt.subplots(figsize=figsize)
# Handle different shap_values formats
if isinstance(self.shap_values, list):
shap_values_plot = self.shap_values[0]
expected_value = self.expected_value[0] if isinstance(self.expected_value, (list, np.ndarray)) else self.expected_value
else:
shap_values_plot = self.shap_values
expected_value = self.expected_value
shap.waterfall_plot(
shap.Explanation(
values=shap_values_plot[instance_idx],
base_values=expected_value,
data=self.X.iloc[instance_idx].values,
feature_names=self.feature_names
),
max_display=max_display,
show=False
)
plt.title(f'SHAP Waterfall Plot - Instance {instance_idx}', fontsize=14, fontweight='bold', pad=20)
plt.tight_layout()
return fig
def plot_dependence(self, feature: str,
interaction_feature: Optional[str] = None,
figsize: Tuple[int, int] = (10, 6)) -> plt.Figure:
"""
Create dependence plot for a feature.
Args:
feature: Feature name to plot
interaction_feature: Feature to use for coloring
figsize: Figure size tuple
Returns:
matplotlib Figure object
"""
fig, ax = plt.subplots(figsize=figsize)
# Handle different shap_values formats
if isinstance(self.shap_values, list):
shap_values_plot = self.shap_values[0]
else:
shap_values_plot = self.shap_values
feature_idx = self.feature_names.index(feature) if feature in self.feature_names else None
if feature_idx is not None:
shap.dependence_plot(
feature_idx,
shap_values_plot,
self.X,
feature_names=self.feature_names,
interaction_index=interaction_feature,
show=False,
ax=ax
)
plt.title(f'SHAP Dependence Plot: {feature}', fontsize=14, fontweight='bold', pad=20)
plt.tight_layout()
return fig
def explain_instance(self, instance_idx: int) -> Dict:
"""
Get explanation for a single instance.
Args:
instance_idx: Index of the instance
Returns:
Dict with explanation details
"""
if isinstance(self.shap_values, list):
shap_values = self.shap_values[0][instance_idx]
expected_value = self.expected_value[0] if isinstance(self.expected_value, (list, np.ndarray)) else self.expected_value
else:
shap_values = self.shap_values[instance_idx]
expected_value = self.expected_value
# Create feature contribution dataframe
contributions = pd.DataFrame({
'feature': self.feature_names,
'value': self.X.iloc[instance_idx].values,
'shap_value': shap_values,
'abs_shap_value': np.abs(shap_values)
}).sort_values('abs_shap_value', ascending=False)
prediction = expected_value + shap_values.sum()
return {
'instance_index': instance_idx,
'expected_value': expected_value,
'prediction': prediction,
'contributions': contributions.to_dict('records')
}
def get_global_explanations(self) -> Dict:
"""
Get global explanations for the model.
Returns:
Dict with global explanation metrics
"""
importance_df = self.get_feature_importance()
return {
'top_features': importance_df.head(10).to_dict('records'),
'feature_count': len(self.feature_names),
'mean_shap_value': np.abs(self.shap_values).mean() if not isinstance(self.shap_values, list) else np.abs(np.array(self.shap_values)).mean()
}
class FallbackExplainability:
"""
Fallback explainability engine when SHAP is not available.
Uses built-in feature importance from models.
"""
def __init__(self, model, X: pd.DataFrame, feature_names: Optional[List[str]] = None):
"""
Initialize fallback explainability.
Args:
model: Trained model
X: Feature data
feature_names: List of feature names
"""
self.model = model
self.X = X.copy() if isinstance(X, pd.DataFrame) else pd.DataFrame(X)
self.feature_names = feature_names or self.X.columns.tolist()
def get_feature_importance(self) -> pd.DataFrame:
"""Get feature importance from model."""
if hasattr(self.model, 'feature_importances_'):
importance = self.model.feature_importances_
elif hasattr(self.model, 'coef_'):
importance = np.abs(self.model.coef_)
if importance.ndim > 1:
importance = importance.mean(axis=0)
else:
# Use permutation importance as fallback
from sklearn.inspection import permutation_importance
perm_importance = permutation_importance(self.model, self.X,
np.zeros(len(self.X)),
n_repeats=5, random_state=42)
importance = perm_importance.importances_mean
importance_df = pd.DataFrame({
'feature': self.feature_names,
'importance': importance
}).sort_values('importance', ascending=False)
return importance_df
def plot_feature_importance(self, max_display: int = 15,
figsize: Tuple[int, int] = (10, 8)) -> plt.Figure:
"""Create bar plot of feature importance."""
importance_df = self.get_feature_importance().head(max_display)
fig, ax = plt.subplots(figsize=figsize)
sns.barplot(data=importance_df, y='feature', x='importance', ax=ax, palette='viridis')
ax.set_title('Feature Importance (Model Built-in)', fontsize=14, fontweight='bold')
ax.set_xlabel('Importance')
ax.set_ylabel('Feature')
plt.tight_layout()
return fig
def create_explainer(model, X: pd.DataFrame, feature_names: Optional[List[str]] = None):
"""
Factory function to create appropriate explainer.
Args:
model: Trained model
X: Feature data
feature_names: List of feature names
Returns:
ExplainabilityEngine or FallbackExplainability instance
"""
if SHAP_AVAILABLE:
try:
return ExplainabilityEngine(model, X, feature_names)
except Exception as e:
st.warning(f"SHAP explainer failed, using fallback: {str(e)}")
return FallbackExplainability(model, X, feature_names)
else:
return FallbackExplainability(model, X, feature_names)
# Streamlit display functions
def display_shap_explanations(explainer, X_sample: pd.DataFrame = None):
"""Display SHAP explanations in Streamlit."""
st.subheader("🔍 Model Explainability")
if not SHAP_AVAILABLE:
st.warning("SHAP is not installed. Using built-in feature importance instead.")
# Feature importance
st.write("### Feature Importance")
fig_importance = explainer.plot_feature_importance()
st.pyplot(fig_importance)
# Summary plot (only for SHAP)
if isinstance(explainer, ExplainabilityEngine):
st.write("### SHAP Summary Plot")
try:
fig_summary = explainer.plot_summary()
st.pyplot(fig_summary)
except Exception as e:
st.warning(f"Could not generate summary plot: {str(e)}")
# Waterfall plot for first instance
if X_sample is not None and len(X_sample) > 0:
st.write("### Individual Prediction Explanation")
try:
fig_waterfall = explainer.plot_waterfall(instance_idx=0)
st.pyplot(fig_waterfall)
except Exception as e:
st.warning(f"Could not generate waterfall plot: {str(e)}")
|