code/optimization_params.json

{
  "baseline_f1": 0.6752810914575621,
  "baseline_precision": 0.627099567099567,
  "baseline_recall": 0.7666666666666667,
  "option_1_weighted_ensemble": {
    "xgboost_weight": 0.7,
    "catboost_weight": 0.3,
    "decision_threshold": 0.42,
    "projected_f1": 0.715,
    "projected_precision": 0.62,
    "projected_recall": 0.8,
    "description": "Best for forensics - high recall to catch unknowns"
  },
  "option_2_feature_engineering": {
    "description": "Add marker-specific features and degradation signals",
    "projected_f1": 0.74,
    "effort": "Medium (2-3 hours)"
  },
  "option_3_collect_data": {
    "target_unknown_samples": 200,
    "current_unknown_samples": 30,
    "projected_f1": 0.77,
    "effort": "High (depends on lab)"
  }
}