FeatureSensitivityAnalyzer

class blackbox2c.analysis.FeatureSensitivityAnalyzer(n_repeats=10, random_state=42)

Measures feature importance via permutation: shuffles each feature's values and measures how much the model's score drops. Features that cause large drops are important.

Constructor

Parameter	Default	Description
n_repeats	10	Number of permutation repetitions per feature
random_state	42	Random seed

analyze()

results = analyzer.analyze(
    model,
    X,
    y,
    feature_names=None,
    scoring=None
)

Parameters:

Parameter	Description
model	Fitted sklearn estimator
X	Input data (n_samples, n_features)
y	Target values
feature_names	Optional list of names
scoring	Scoring function (default: auto-detect accuracy or R2)

Returns: SensitivityResults

---

SensitivityResults

Methods

results.summary()
# Returns a formatted string with importances for all features

results.get_top_features(n)
# Returns list of (index, name, importance) for the top-N features

results.get_optimal_subset(threshold=0.01, min_features=1)
# Returns list of feature indices above the importance threshold

results.get_redundant_features(threshold=0.001)
# Returns list of feature indices with importance near zero

Plotting

fig, ax = results.plot(figsize=(10, 6), save_path=None)
# Returns (fig, ax). Requires matplotlib.
# Displays a horizontal bar chart with permutation importances and error bars.

---

Example

from sklearn.ensemble import RandomForestClassifier
from sklearn.datasets import load_iris
from blackbox2c.analysis import FeatureSensitivityAnalyzer

iris = load_iris()
model = RandomForestClassifier(n_estimators=50, random_state=42)
model.fit(iris.data, iris.target)

analyzer = FeatureSensitivityAnalyzer(n_repeats=20, random_state=42)
results = analyzer.analyze(
    model, iris.data, iris.target,
    feature_names=list(iris.feature_names)
)

print(results.summary())
top2 = results.get_top_features(2)
print("Top 2 features:", top2)