anomaly-grid-py 0.3.0

High-performance sequence anomaly detection with minimal dependencies

🔍 Anomaly Grid Python

High-performance sequence anomaly detection using Markov models. Minimal dependencies, scikit-learn style API.

Installation

pip install anomaly-grid-py

Quick Start

from anomaly_grid_py import AnomalyDetector

# Create detector
detector = AnomalyDetector(max_order=3)

# Train on normal sequences
normal_patterns = [
    ['LOGIN', 'BALANCE', 'LOGOUT'],
    ['LOGIN', 'WITHDRAW', 'LOGOUT'],
    ['LOGIN', 'TRANSFER', 'LOGOUT']
] * 100

detector.fit(normal_patterns)

# Detect anomalies
test_sequences = [
    ['LOGIN', 'BALANCE', 'LOGOUT'],      # Normal
    ['HACK', 'EXPLOIT', 'STEAL'],        # Anomalous
]

# Get anomaly scores (0-1, higher = more anomalous)
scores = detector.predict_proba(test_sequences)
print(f"Scores: {scores}")

# Get binary predictions
anomalies = detector.predict(test_sequences, threshold=0.1)
print(f"Anomalies: {anomalies}")

API Reference

AnomalyDetector(max_order=3)

Methods:

• fit(X) - Train on list of sequences
• predict_proba(X) - Get anomaly scores (0-1)
• predict(X, threshold=0.1) - Get binary predictions
• get_performance_metrics() - Get training metrics

Parameters:

• X: List of sequences, each sequence is a list of strings
• threshold: Detection threshold (0-1)

Examples

See example.py for complete examples:

python example.py

License

MIT License - see LICENSE file for details.