detect.py

#!/usr/bin/env python3
import argparse
from pathlib import Path

from src.detector import run_detection_pipeline
from src.explanations import anomaly_summary
from src.utils import setup_logger, ensure_dirs

logger = setup_logger("detect")

PROCESSED_DIR = Path("data/processed")
ANOMALIES_DIR = Path("data/anomalies")
MODEL_DIR = Path("models")
ensure_dirs([PROCESSED_DIR, ANOMALIES_DIR, MODEL_DIR])


def main():
    parser = argparse.ArgumentParser(description="Run anomaly detection on processed features")
    parser.add_argument("--input", type=str, default="features.csv", help="Processed features CSV")
    parser.add_argument("--output", type=str, default="anomaly_events.csv", help="Anomaly output CSV")
    parser.add_argument("--contamination", type=float, default=0.1, help="Expected anomaly proportion")
    args = parser.parse_args()

    features_path = PROCESSED_DIR / args.input
    output_path = ANOMALIES_DIR / args.output

    if not features_path.exists():
        logger.error("Features not found at %s", features_path)
        logger.info("Run `python train_baseline.py` first")
        return

    anomalies = run_detection_pipeline(features_path, MODEL_DIR, output_path, contamination=args.contamination)
    print()
    print("=" * 60)
    print("ANOMALY DETECTION RESULTS")
    print("=" * 60)
    print(anomaly_summary(anomalies))
    print("=" * 60)
    print(f"Full results saved to {output_path}")


if __name__ == "__main__":
    main()