Phishing-Detection-System / scripts /testing /test_feature_alignment.py
rb1337's picture
Upload 50 files
2cc7f91 verified
Raw History Blame Contribute Delete
4.52 kB
"""
Test feature alignment between extractor and models
"""
import sys
from pathlib import Path
import joblib
import pandas as pd
sys.path.append(str(Path(__file__).parent))
from scripts.feature_extraction.url_features_v2 import URLFeatureExtractorV2
def test_feature_alignment():
"""Test that feature extraction produces features in the correct order for models"""
# Load models
models_dir = Path(__file__).parent / 'saved_models'
model_files = {
'Logistic Regression': 'logistic_regression.joblib',
'Random Forest': 'random_forest.joblib',
'XGBoost': 'xgboost.joblib'
}
# Load scaler
scaler_path = models_dir / 'scaler.joblib'
scaler = None
if scaler_path.exists():
scaler = joblib.load(scaler_path)
print(f"✓ Loaded scaler")
if hasattr(scaler, 'feature_names_in_'):
print(f" Scaler has {len(scaler.feature_names_in_)} feature names\n")
# Initialize extractor
extractor = URLFeatureExtractorV2()
# Test URL
test_url = "https://github.com/user/repo"
print("Testing feature alignment...\n")
print(f"Test URL: {test_url}\n")
# Extract features
features_dict = extractor.extract_features(test_url)
features_df = pd.DataFrame([features_dict])
if 'label' in features_df.columns:
features_df = features_df.drop('label', axis=1)
print(f"Extracted {len(features_df.columns)} features\n")
# Store feature names for fallback
feature_names_store = {}
# Check each model
for name, filename in model_files.items():
model_path = models_dir / filename
if not model_path.exists():
print(f"❌ {name}: Model file not found")
continue
model = joblib.load(model_path)
# Determine expected features
expected_features = None
source = None
if hasattr(model, 'feature_names_in_'):
expected_features = list(model.feature_names_in_)
source = "model"
elif hasattr(scaler, 'feature_names_in_'):
expected_features = list(scaler.feature_names_in_)
source = "scaler"
elif feature_names_store:
expected_features = list(feature_names_store.values())[0]
source = "fallback"
if expected_features:
feature_names_store[name] = expected_features
print(f"✓ {name}:")
print(f" Expected features: {len(expected_features)} (from {source})")
print(f" Expected features: {len(expected_features)} (from {source})")
# Check missing features
missing = set(expected_features) - set(features_df.columns)
extra = set(features_df.columns) - set(expected_features)
if missing:
print(f" ⚠ Missing features: {len(missing)}")
print(f" {list(missing)[:5]}...")
if extra:
print(f" ⚠ Extra features: {len(extra)}")
print(f" {list(extra)[:5]}...")
if not missing and not extra:
print(f" ✓ Perfect match!")
# Try prediction with alignment
features_aligned = pd.DataFrame(columns=expected_features)
for feat in expected_features:
if feat in features_df.columns:
features_aligned[feat] = features_df[feat].values
else:
features_aligned[feat] = 0
# Scale for Logistic Regression
if name == 'Logistic Regression' and scaler is not None:
features_to_use = scaler.transform(features_aligned)
else:
features_to_use = features_aligned
try:
pred = model.predict(features_to_use)[0]
proba = model.predict_proba(features_to_use)[0]
print(f" ✓ Prediction successful: {'PHISHING' if pred == 1 else 'LEGITIMATE'} ({proba[pred]*100:.1f}%)")
except Exception as e:
print(f" ❌ Prediction failed: {e}")
else:
print(f"⚠ {name}: No feature names available")
print()
if __name__ == "__main__":
test_feature_alignment()