# TRIPOD+AI record

> Generated by `matrixai report --tripod` from the package manifest. Every line comes from what the run recorded; what is not known is said.
>
> This is not a medical device nor a regulatory approval, and it does not replace the judgement of whoever signs the study.
>
> **WARNING: the training data is SYNTHETIC**, generated by the recipe shipped in this package. A model trained this way is not clinically validated.

## Data
- **Source**: `coherent` · `reproduce.json#generation.mode`
- **Rows that trained**: 400 · `reproduce.json#artifacts.dataset.rows`
- **Dataset digest (sha256)**: 1cefbf74e237efdae2f6e06d7b9c5141ed815cccbd547aa18341ab7a79c96e65 · `reproduce.json#artifacts.dataset.sha256`
- **Predictors (inputs)**: edad (18–100), dias_ingresado (0–60), num_diagnosticos (0–15), reingreso_previo (0–1) · `model.mxai#vectors`
- **Excluded columns**: _not available_ · `reproduce.json#generation.excluded_identifiers`
- **Split and seeds**: {"dataset": 20260825, "init": 42, "split": 42} · `reproduce.json#generation.seeds`

**Data recipe** · `data_recipe.txt#`:

```
alto: edad > 75 OR reingreso_previo > 0.5
DEFAULT: bajo
```

## Model
- **Architecture (`.mxai`)**: a164f4380120be24d6cc66ac564bc7af820498cdbcf0caf251398aa904410747 · `reproduce.json#artifacts.model.sha256`
- **Hyperparameters (`.mxtrain`)**: 45235ce15bb7744a43dc7aee32ef1622009be352e2d0926e0d3bd8d7feca18d6 · `reproduce.json#artifacts.training.sha256`
- **Epochs declared / effective / ran**: 60 / _not available_ / _not available_ · `reproduce.json#generation.epochs_declared+effective+ran`
- **Environment**: matrixai 1.8.0 · python 3.12.3 · stdlib/cpu · `reproduce.json#environment`

## Performance
- **accuracy**: 0.96875 (measured on dataset `1cefbf74e237…`) · `reproduce.json#metrics`
- **best_validation_loss**: 0.07562131317942544 (measured on dataset `1cefbf74e237…`) · `reproduce.json#metrics`

## Model availability · Reproducibility
- The package ships `model.onnx`, `predict.py` and `inference_spec.json`: it predicts from raw values with no MatrixAI installed.
- The package declares itself reproducible. Check it with `matrixai verify <package>`. · `reproduce.json#reproducible`

## What this record CANNOT fill in
- **Missing data**: this package declares no missing-value policy, so what was done with them cannot be stated · `clinical_profile.json#politica_de_faltantes`
- **Calibration**: this package ships no calibration curve, which the core does know how to measure (105-C3) · `clinical_profile.json#calibracion.ece`
- **Subgroup fairness**: this package ships no measured subgroup; it needs a team to predefine them (105-C4) · `clinical_profile.json#segmentos`
- **Funding**: belongs to a study's author, not to a run
- **Conflicts of interest**: idem
- **Ethical approval and study registration**: idem
- **Intended use and target population**: a person writes this; this report asks for it, it does not invent it · `team_declaration.json#uso_previsto`