w60s5_binary/models/ExtraTrees_metadata.json

{
  "model_id": "extratrees_chest_ecg_w60s5_binary_v1_0",
  "format": "onnx",
  "task": "binary_classification",
  "window_config": {
    "window_size_sec": 60,
    "window_step_sec": 5
  },
  "schema": {
    "input_names": [
      "RMSSD",
      "Mean_RR",
      "HRV_SDNN",
      "pNN50",
      "HRV_HF",
      "HRV_LF",
      "HRV_HF_nu",
      "HRV_LF_nu",
      "HRV_LFHF",
      "HRV_TP",
      "HRV_SD1SD2",
      "HRV_Sampen",
      "HRV_DFA_alpha1",
      "HR"
    ],
    "input_units": [
      "ms",
      "ms",
      "ms",
      "ms",
      "ms\u00b2",
      "ms\u00b2",
      "ms\u00b2",
      "ms\u00b2",
      "ms\u00b2",
      "ms\u00b2",
      "ratio",
      "entropy",
      "alpha",
      "bpm"
    ],
    "order_fixed": true,
    "normalization": {
      "type": "standard_scaler",
      "mean": [
        52.52255488912746,
        795.1433169932258,
        69.2944695053504,
        23.194129727420158,
        0.0,
        0.0,
        0.0,
        0.0,
        0.0,
        0.0,
        0.0,
        0.0,
        0.0,
        78.58138116480158
      ],
      "std": [
        37.378859331087114,
        149.78295254278115,
        30.15453668337926,
        19.105756263894023,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        1.0,
        17.32690899339997
      ],
      "note": "Normalization included in the ONNX model pipeline"
    }
  },
  "output": {
    "type": "probability",
    "range": [
      0.0,
      1.0
    ],
    "class_names": [
      "Baseline",
      "Stress"
    ],
    "class_indices": [
      0,
      1
    ],
    "positive_class": "Stress"
  },
  "quantization": {
    "enabled": false,
    "dtype": "float32"
  },
  "performance": {
    "loso_cv_accuracy": 0.7670234570712822,
    "loso_cv_f1_score": 0.7040786809372288
  },
  "training_data_tag": "wesad_chest_ecg_binary",
  "created_utc": "2025-12-26T16:21:12.326652Z",
  "license": "Apache-2.0",
  "notes": "Binary classification (Baseline vs Stress). ONNX model with built-in StandardScaler normalization. Trained on WESAD chest ECG with 60s windows and 5s steps.",
  "checksum": {
    "algo": "sha256",
    "value": "b34f011c436b8d1ea122071175def85b7bc6f1458753d8114a2da55ef9de6c73"
  }
}