{
  "model_name": "isolation_forest_quality",
  "model_type": "IsolationForest (NumPy implementation)",
  "implementation_module": "isolation_forest.py (in model/ dir)",
  "version": "1.0.0",
  "training_date": "2026-08-13",
  "operational_date": "2026-08-13",
  "dataset": "toolathon123/project_20260813_014231_9dbe9212",
  "n_records_trained": 2000,
  "n_estimators": 100,
  "max_samples": 256,
  "contamination": 0.05,
  "features": [
    "delay_minutes",
    "damage_flag",
    "temperature_deviation",
    "humidity_deviation",
    "fuel_consumption",
    "quality_score"
  ],
  "anomaly_threshold": 0.6043743324007921,
  "metrics": {
    "high_risk_ratio": 0.05,
    "n_high_risk": 100,
    "avg_delay_minutes_all": 83.19,
    "avg_delay_minutes_low_risk": 77.43,
    "avg_delay_minutes_high_risk": 192.71,
    "delay_improvement_rate_pct": 6.9286,
    "damage_rate_pct": 8.1,
    "avg_quality_score": 85.52,
    "avg_temperature_deviation": 0.63,
    "avg_humidity_deviation": 4.566
  },
  "output_meaning": {
    "anomaly_score": "Isolation Forest anomaly score in [0.5, 1.0]; higher = more anomalous",
    "high_risk": "1 = high-quality-risk shipment flagged for review, 0 = normal"
  },
  "update_strategy": "Retrain daily on each day's new data snapshot; threshold re-estimated at 5% contamination."
}