Files
defrag/app/classification/ml.py
2025-12-13 11:56:06 +01:00

128 lines
4.8 KiB
Python

from pathlib import Path
from typing import Optional, List, Tuple
import pickle
try:
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.naive_bayes import MultinomialNB
from sklearn.pipeline import Pipeline
SKLEARN_AVAILABLE = True
except ImportError:
SKLEARN_AVAILABLE = False
class MLClassifier:
def __init__(self):
if not SKLEARN_AVAILABLE:
raise ImportError('scikit-learn is required for ML classification. Install with: pip install scikit-learn')
self.model: Optional[Pipeline] = None
self.categories: List[str] = []
self._is_trained = False
def _extract_features(self, path: Path) -> str:
parts = path.parts
extension = path.suffix
filename = path.name
features = []
features.extend(parts)
if extension:
features.append(f'ext:{extension}')
name_parts = filename.replace('-', ' ').replace('_', ' ').replace('.', ' ').split()
features.extend([f'name:{part}' for part in name_parts])
return ' '.join(features)
def train(self, training_data: List[Tuple[Path, str]]) -> None:
if not training_data:
raise ValueError('Training data cannot be empty')
X = [self._extract_features(path) for path, _ in training_data]
y = [category for _, category in training_data]
self.categories = sorted(set(y))
self.model = Pipeline([('tfidf', TfidfVectorizer(max_features=1000, ngram_range=(1, 2), min_df=1)), ('classifier', MultinomialNB())])
self.model.fit(X, y)
self._is_trained = True
def classify(self, path: Path, file_type: Optional[str]=None) -> Optional[str]:
if not self._is_trained or self.model is None:
return None
features = self._extract_features(path)
try:
prediction = self.model.predict([features])[0]
return prediction
except Exception:
return None
def predict_proba(self, path: Path) -> dict[str, float]:
if not self._is_trained or self.model is None:
return {}
features = self._extract_features(path)
try:
probabilities = self.model.predict_proba([features])[0]
return {category: float(prob) for category, prob in zip(self.categories, probabilities)}
except Exception:
return {}
def save_model(self, model_path: Path) -> None:
if not self._is_trained:
raise ValueError('Cannot save untrained model')
model_data = {'model': self.model, 'categories': self.categories, 'is_trained': self._is_trained}
with open(model_path, 'wb') as f:
pickle.dump(model_data, f)
def load_model(self, model_path: Path) -> None:
with open(model_path, 'rb') as f:
model_data = pickle.load(f)
self.model = model_data['model']
self.categories = model_data['categories']
self._is_trained = model_data['is_trained']
@property
def is_trained(self) -> bool:
return self._is_trained
class DummyMLClassifier:
def __init__(self):
pass
def train(self, training_data: List[Tuple[Path, str]]) -> None:
raise NotImplementedError('ML classification requires scikit-learn. Install with: pip install scikit-learn')
def classify(self, path: Path, file_type: Optional[str]=None) -> Optional[str]:
return None
def predict_proba(self, path: Path) -> dict[str, float]:
return {}
def save_model(self, model_path: Path) -> None:
raise NotImplementedError('ML classification not available')
def load_model(self, model_path: Path) -> None:
raise NotImplementedError('ML classification not available')
@property
def is_trained(self) -> bool:
return False
def create_ml_classifier() -> MLClassifier | DummyMLClassifier:
if SKLEARN_AVAILABLE:
return MLClassifier()
else:
return DummyMLClassifier()
def train_from_database(db_connection, min_samples_per_category: int=10) -> MLClassifier | DummyMLClassifier:
classifier = create_ml_classifier()
if isinstance(classifier, DummyMLClassifier):
return classifier
cursor = db_connection.cursor()
cursor.execute('\n SELECT path, category\n FROM files\n WHERE category IS NOT NULL\n ')
training_data = [(Path(path), category) for path, category in cursor.fetchall()]
cursor.close()
if not training_data:
return classifier
category_counts = {}
for _, category in training_data:
category_counts[category] = category_counts.get(category, 0) + 1
filtered_data = [(path, category) for path, category in training_data if category_counts[category] >= min_samples_per_category]
if filtered_data:
classifier.train(filtered_data)
return classifier