ml/artifacts/figures/comparison/comparison_report.json

{
  "weights_used": {
    "macro_f1": 0.5,
    "accuracy": 0.2,
    "latency": 0.15,
    "model_size": 0.1,
    "calibration": 0.05
  },
  "rankings": [
    {
      "model_name": "resnet",
      "composite_score": 0.9241,
      "raw_metrics": {
        "macro_f1": 0.5606227821703,
        "accuracy": 0.6318082788671024,
        "latency_ms": 5.68,
        "model_size_mb": 93.73
      },
      "normalized_metrics": {
        "macro_f1": 1.0,
        "accuracy": 1.0,
        "latency": 0.7485,
        "model_size": 0.8685
      },
      "rank": 1
    },
    {
      "model_name": "vit",
      "composite_score": 0.7542,
      "raw_metrics": {
        "macro_f1": 0.5465106856964207,
        "accuracy": 0.6318082788671024,
        "latency_ms": 15.5,
        "model_size_mb": 328.85
      },
      "normalized_metrics": {
        "macro_f1": 0.967,
        "accuracy": 1.0,
        "latency": 0.0,
        "model_size": 0.4575
      },
      "rank": 2
    },
    {
      "model_name": "cnn",
      "composite_score": 0.2898,
      "raw_metrics": {
        "macro_f1": 0.14429494847203667,
        "accuracy": 0.2440087145969499,
        "latency_ms": 2.38,
        "model_size_mb": 18.52
      },
      "normalized_metrics": {
        "macro_f1": 0.0252,
        "accuracy": 0.0111,
        "latency": 1.0,
        "model_size": 1.0
      },
      "rank": 3
    },
    {
      "model_name": "mlp",
      "composite_score": 0.1731,
      "raw_metrics": {
        "macro_f1": 0.1335123629487021,
        "accuracy": 0.23965141612200436,
        "latency_ms": 2.55,
        "model_size_mb": 590.55
      },
      "normalized_metrics": {
        "macro_f1": 0.0,
        "accuracy": 0.0,
        "latency": 0.987,
        "model_size": 0.0
      },
      "rank": 4
    }
  ],
  "best_model": "resnet",
  "recommendation": "Recommended model: resnet (composite score: 0.9241)"
}