Dataset titanic

Feature Scatter

Medical treatment outcomes derived from the Titanic passenger data.

Provenance

{
  "metadata": {
    "id": 1,
    "primary_key": "PassengerId",
    "target_variable": "Survived",
    "original_dataset_type": "passenger survival",
    "obfuscated_dataset_type": "medical treatment",
    "obfuscated_table_name": "patients",
    "obfuscated_split_table_name": "patient_splits",
    "created_at": "2025-03-09 04:33:42"
  },
  "column_transformations": [
    {
      "original_column": "Survived",
      "remove": false,
      "obfuscated_column": "Outcome",
      "transformation": "obfuscated['Outcome'] = original['Survived'].map({0: 'Failure', 1: 'Success'})"
    },
    {
      "original_column": "PassengerId",
      "remove": false,
      "obfuscated_column": "Patient_ID",
      "transformation": "obfuscated['Patient_ID'] = original['PassengerId']"
    },
    {
      "original_column": "Pclass",
      "remove": false,
      "obfuscated_column": "Histogen_Complex",
      "transformation": "obfuscated['Histogen_Complex'] = original['Pclass'].map({1: 'Beta', 2: 'Omicron', 3: 'Delta'})"
    },
    {
      "original_column": "Name",
      "remove": true,
      "obfuscated_column": null,
      "transformation": null
    },
    {
      "original_column": "Sex",
      "remove": false,
      "obfuscated_column": "Sex",
      "transformation": "obfuscated['Sex'] = original['Sex'].map({'male': 'female', 'female': 'male'})"
    },
    {
      "original_column": "Age",
      "remove": false,
      "obfuscated_column": "Treatment_Months",
      "transformation": "obfuscated['Treatment_Months'] = original['Age'].fillna(original['Age'].mean()) * 3"
    },
    {
      "original_column": "SibSp",
      "remove": false,
      "obfuscated_column": "Genetic_Class_A_Matches",
      "transformation": "obfuscated['Genetic_Class_A_Matches'] = original['SibSp'] + 1"
    },
    {
      "original_column": "Parch",
      "remove": false,
      "obfuscated_column": "Genetic_Class_B_Matches",
      "transformation": "obfuscated['Genetic_Class_B_Matches'] = original['Parch'] + 1"
    },
    {
      "original_column": "Fare",
      "remove": false,
      "obfuscated_column": "TcQ_mass",
      "transformation": "obfuscated['TcQ_mass'] = original['Fare'] * 1000"
    },
    {
      "original_column": "Cabin",
      "remove": true,
      "obfuscated_column": null,
      "transformation": null
    },
    {
      "original_column": "Embarked",
      "remove": false,
      "obfuscated_column": "Cohort",
      "transformation": "obfuscated['Cohort'] = original['Embarked'].map({'S': 'Melbourne', 'C': 'Delhi', 'Q': 'Lisbon'})"
    }
  ]
}

Best Validation Round Rank

Test scores by release date

Regression slope: 0.0006, intercept: -12.0403, p-value: 0.0051174

Debug

merge info: 21 rows
filter ollama: 20 rows
drop NA release/accuracy: 20 rows
compute KT and sort
added baselines
create figure
scatter models: [(datetime.date(2024, 1, 25), -0.29260619125154674), (datetime.date(2024, 7, 18), -0.16363430424699762), (datetime.date(2024, 10, 22), -0.2025191059757678), (datetime.date(2024, 10, 22), -0.2025191059757678), (datetime.date(2024, 11, 20), -0.2162948722980533), (datetime.date(2024, 11, 20), -0.26045561811203355), (datetime.date(2024, 12, 5), -0.15140970440866616), (datetime.date(2024, 12, 5), -0.12794676600021637), (datetime.date(2025, 1, 30), -0.17621299542141272), (datetime.date(2025, 1, 30), -0.2452309432186094), (datetime.date(2025, 2, 19), -0.17621299542141272), (datetime.date(2025, 2, 19), -0.09497042173419899), (datetime.date(2025, 4, 14), -0.18916690423326227), (datetime.date(2025, 4, 14), -0.15140970440866616), (datetime.date(2025, 4, 16), -0.16363430424699762), (datetime.date(2025, 4, 16), -0.10568671363512454), (datetime.date(2025, 5, 14), -0.13951980510549647), (datetime.date(2025, 5, 14), -0.2162948722980533), (datetime.date(2025, 5, 14), -0.10568671363512454), (datetime.date(2025, 5, 14), -0.12794676600021637)]
xlim_min 2024-01-01
scatter ensembles: 4
draw ensemble trend line
xlim_max 2025-05-14
axhline ebm at -0.116
annotate ebm at edge 20222.0
mark baseline crossing
savefig website/dataset/titanic/release_scores_titanic.png

Model Scores

Train vs Validation

Regression slope: 0.7902, intercept: 0.2046, p-value: 0.0034905

Validation vs Train

Regression slope: 0.4877, intercept: 0.3092, p-value: 0.0034905

Ensemble Max Scores

Baseline

Baseline Models

Logistic Regression

take the cohort_Delhi and multiply by 0.410, take the cohort_Lisbon and multiply by 0.380, take the cohort_Melbourne and multiply by -0.099, take the cohort_NaN and multiply by 0.103, take the genetic_class_a_matches and multiply by -0.286, take the genetic_class_b_matches and multiply by -0.094, take the histogen_complex_Beta and multiply by 1.216, take the histogen_complex_Delta and multiply by -0.741, take the histogen_complex_Omicron and multiply by 0.320, take the sex_female and multiply by -0.902, take the sex_male and multiply by 1.696, take the tcq_mass and multiply by 0.000, take the treatment_months and multiply by -0.013, add those up and if the number is greater than 0.998 then predict Failure, otherwise predict Success.

Decision Tree

Dummy

Always predict 0

RuleFit

Bayesian Rule List

CORELS

1. {'col': 'X4', 'index_col': 4, 'cutoff': 0.5, 'val': 0.1872146118721461, 'flip': False, 'val_right': 0.7372549019607844, 'num_pts': 693, 'num_pts_right': 255}
2. {'col': 'X0', 'index_col': 0, 'cutoff': 0.5, 'val': 0.13872832369942195, 'flip': False, 'val_right': 0.3695652173913043, 'num_pts': 438, 'num_pts_right': 92}
3. {'col': 'X5', 'index_col': 5, 'cutoff': 0.5, 'val': 0.10218978102189781, 'flip': False, 'val_right': 0.2777777777777778, 'num_pts': 346, 'num_pts_right': 72}
4. {'col': 'X7', 'index_col': 7, 'cutoff': 0.5, 'val': 0.0946969696969697, 'flip': False, 'val_right': 0.3, 'num_pts': 274, 'num_pts_right': 10}
5. {'col': 'X19', 'index_col': 19, 'cutoff': 0.5, 'val': 0.08583690987124463, 'flip': False, 'val_right': 0.16129032258064516, 'num_pts': 264, 'num_pts_right': 31}
6. {'val': 0.08583690987124463, 'num_pts': 233}

EBM