signaux-faibles / predictsignauxfaibles

Dépôt du code python permettant la production de liste de prédiction Signaux Faibles.
MIT License
6 stars 2 forks source link

F_\beta score is null using default CLI parameters #81

Open raphaelventura opened 3 years ago

raphaelventura commented 3 years ago

When launching the CLI tool using default parameters (with either develop or prod as values for the ENV variable), the output F_\beta score is always null. See, e.g. this sample output:

{
  "train_spl_size": 1000000,
  "test_spl_size": 250000,
  "predict_spl_size": 1000000000,
  "train_proportion_positive_class": 0.2,
  "train_from": "2016-01-01",
  "train_to": "2018-06-30",
  "test_from": "2018-07-01",
  "test_to": "2018-10-31",
  "predict_on": "2020-02-01",
  "run_on": "20210827-115657",
  "train": {
    "balanced_accuracy": 0.720742396545498,
    "Fbeta": null
  },
  "test": {
    "balanced_accuracy": 0.5670635380514673,
    "Fbeta": null
  },
  "predict": {
  }
}

We should investigate where this comes from and create an associated unit test in order to avoid regression.