Dataset southgermancredit

Feature Scatter

Coral reef health metrics based on the South German Credit dataset.

Provenance

{
  "metadata": {
    "id": 1,
    "primary_key": "k_id",
    "target_variable": "kredit",
    "original_dataset_type": "credit card",
    "obfuscated_dataset_type": "coral reef health",
    "obfuscated_table_name": "coral_reefs",
    "obfuscated_split_table_name": "coral_reef_splits",
    "created_at": "2025-03-17 17:40:58"
  },
  "column_transformations": [
    {
      "original_column": "k_id",
      "remove": false,
      "obfuscated_column": "CoralReefID",
      "transformation": null
    },
    {
      "original_column": "laufkont",
      "remove": false,
      "obfuscated_column": "CurrentFlowQuality",
      "transformation": "obfuscated['CurrentFlowQuality'] = original['laufkont'].map({1: 'Poor', 2: 'Fair', 3: 'Good', 4: 'Excellent'})"
    },
    {
      "original_column": "laufzeit",
      "remove": false,
      "obfuscated_column": "ObservationDuration",
      "transformation": "obfuscated['ObservationDuration'] = original['laufzeit'] * 3 + 7"
    },
    {
      "original_column": "moral",
      "remove": false,
      "obfuscated_column": "ReefIntegrityScore",
      "transformation": "obfuscated['ReefIntegrityScore'] = original['moral'].map({0: 'Very Poor', 1: 'Poor', 2: 'Moderate', 3: 'Good', 4: 'Excellent'})"
    },
    {
      "original_column": "verw",
      "remove": false,
      "obfuscated_column": "PredatorActivityLevel",
      "transformation": "obfuscated['PredatorActivityLevel'] = original['verw'].map({0: 'None', 1: 'Very Low', 2: 'Low', 3: 'Moderate Low', 4: 'Moderate', 5: 'Moderate High', 6: 'High', 7: 'Very High', 8: 'Extreme', 9: 'Severe', 10: 'Catastrophic'})"
    },
    {
      "original_column": "hoehe",
      "remove": false,
      "obfuscated_column": "AcousticIntensity",
      "transformation": "obfuscated['AcousticIntensity'] = original['hoehe'] / 10 + 2"
    },
    {
      "original_column": "sparkont",
      "remove": false,
      "obfuscated_column": "AlgalCoverage",
      "transformation": "obfuscated['AlgalCoverage'] = original['sparkont'].map({1: 'Minimal', 2: 'Sparse', 3: 'Moderate', 4: 'Heavy', 5: 'Extensive'})"
    },
    {
      "original_column": "beszeit",
      "remove": false,
      "obfuscated_column": "CoralAgeEstimate",
      "transformation": "obfuscated['CoralAgeEstimate'] = original['beszeit'] * 5 + 1"
    },
    {
      "original_column": "rate",
      "remove": false,
      "obfuscated_column": "BleachingEventsPerYear",
      "transformation": "obfuscated['BleachingEventsPerYear'] = original['rate'].map({1: 'Rare', 2: 'Occasional', 3: 'Frequent', 4: 'Constant'})"
    },
    {
      "original_column": "famges",
      "remove": false,
      "obfuscated_column": "BiodiversityIndex",
      "transformation": "obfuscated['BiodiversityIndex'] = original['famges'].map({1: 'Low', 2: 'Medium', 3: 'High', 4: 'Very High'})"
    },
    {
      "original_column": "buerge",
      "remove": false,
      "obfuscated_column": "NearbyHealthyReef",
      "transformation": "obfuscated['NearbyHealthyReef'] = original['buerge'].map({1: 'Absent', 2: 'Nearby', 3: 'Adjacent'})"
    },
    {
      "original_column": "wohnzeit",
      "remove": false,
      "obfuscated_column": "ReefMonitoringDuration",
      "transformation": "obfuscated['ReefMonitoringDuration'] = original['wohnzeit'].map({1: '< 1 year', 2: '1-3 years', 3: '4-6 years', 4: '> 6 years'})"
    },
    {
      "original_column": "verm",
      "remove": false,
      "obfuscated_column": "PollutionLevel",
      "transformation": "obfuscated['PollutionLevel'] = original['verm'].map({1: 'Low', 2: 'Moderate', 3: 'High', 4: 'Critical'})"
    },
    {
      "original_column": "alter",
      "remove": false,
      "obfuscated_column": "ReefAverageAge",
      "transformation": "obfuscated['ReefAverageAge'] = original['alter'] + 3"
    },
    {
      "original_column": "weitkred",
      "remove": false,
      "obfuscated_column": "DistantStressIndicators",
      "transformation": "obfuscated['DistantStressIndicators'] = original['weitkred'].map({1: 'None', 2: 'Possible', 3: 'Confirmed'})"
    },
    {
      "original_column": "wohn",
      "remove": false,
      "obfuscated_column": "ReefDepthZone",
      "transformation": "obfuscated['ReefDepthZone'] = original['wohn'].map({1: 'Shallow', 2: 'Mid-depth', 3: 'Deep'})"
    },
    {
      "original_column": "bishkred",
      "remove": false,
      "obfuscated_column": "PreviousStressIncidents",
      "transformation": "obfuscated['PreviousStressIncidents'] = original['bishkred'].map({1: 'None', 2: 'Minor', 3: 'Moderate', 4: 'Major'})"
    },
    {
      "original_column": "beruf",
      "remove": false,
      "obfuscated_column": "CoralDominantType",
      "transformation": "obfuscated['CoralDominantType'] = original['beruf'].map({1: 'Branching', 2: 'Massive', 3: 'Encrusting', 4: 'Soft'})"
    },
    {
      "original_column": "pers",
      "remove": false,
      "obfuscated_column": "SurveyorExperience",
      "transformation": "obfuscated['SurveyorExperience'] = original['pers'].map({1: 'No', 2: 'Yes'})"
    },
    {
      "original_column": "telef",
      "remove": false,
      "obfuscated_column": "RemoteSensorPresent",
      "transformation": "obfuscated['RemoteSensorPresent'] = original['telef'].map({1: 'No', 2: 'Yes'})"
    },
    {
      "original_column": "gastarb",
      "remove": false,
      "obfuscated_column": "InvasiveSpeciesDetected",
      "transformation": "obfuscated['InvasiveSpeciesDetected'] = original['gastarb'].map({1: 'No', 2: 'Yes'})"
    },
    {
      "original_column": "kredit",
      "remove": false,
      "obfuscated_column": "ReefHealthStatus",
      "transformation": "obfuscated['ReefHealthStatus'] = original['kredit'].map({0: 'Stressed', 1: 'Healthy'})"
    }
  ]
}

Best Validation Round Rank

Test scores by release date

Regression slope: -0.0001, intercept: 2.2000, p-value: 0.18062

Debug

merge info: 29 rows
filter ollama: 26 rows
drop NA release/accuracy: 26 rows
compute KT and sort
added baselines
create figure
scatter models: [(datetime.date(2024, 1, 25), -0.10875110148918317), (datetime.date(2024, 7, 18), -0.2677806157634906), (datetime.date(2024, 10, 22), -0.28224481615501795), (datetime.date(2024, 10, 22), -0.3454614810151819), (datetime.date(2024, 12, 5), -0.26072524332643304), (datetime.date(2024, 12, 5), -0.35405204092877696), (datetime.date(2025, 1, 29), -0.28966167059874204), (datetime.date(2025, 1, 29), -0.29720739362252735), (datetime.date(2025, 1, 30), -0.2677806157634906), (datetime.date(2025, 1, 30), -0.3370375558407868), (datetime.date(2025, 2, 5), -0.28224481615501795), (datetime.date(2025, 2, 19), -0.2537826597343463), (datetime.date(2025, 2, 19), -0.2537826597343463), (datetime.date(2025, 2, 27), -0.28966167059874204), (datetime.date(2025, 2, 27), -0.2335969620668543), (datetime.date(2025, 3, 25), -0.17812421223295416), (datetime.date(2025, 3, 25), -0.4730836765961299), (datetime.date(2025, 4, 14), -0.21430795779792694), (datetime.date(2025, 4, 14), -0.48464812331575086), (datetime.date(2025, 4, 16), -0.3048865427389483), (datetime.date(2025, 4, 16), -0.2206429104634819), (datetime.date(2025, 5, 14), -0.18985365882318037), (datetime.date(2025, 5, 14), -0.10384689799321975), (datetime.date(2025, 5, 14), -0.19583935764749427), (datetime.date(2025, 5, 14), -0.2469493155174297), (datetime.date(2025, 5, 14), -0.13939953213276507)]
xlim_min 2024-01-01
scatter ensembles: 3
draw ensemble trend line
xlim_max 2025-05-14
axhline rulefit at -0.149
annotate rulefit at edge 19723.0
savefig website/dataset/southgermancredit/release_scores_southgermancredit.png

Model Scores

Train vs Validation

Regression slope: 0.7228, intercept: 0.1704, p-value: 2.4544e-06

Validation vs Train

Regression slope: 0.8446, intercept: 0.0735, p-value: 2.4544e-06

Ensemble Max Scores

Baseline

Baseline Models

Logistic Regression

take the acousticintensity and multiply by 0.001, take the algalcoverage_Extensive and multiply by -0.625, take the algalcoverage_Heavy and multiply by -0.424, take the algalcoverage_Minimal and multiply by 0.474, take the algalcoverage_Moderate and multiply by -0.016, take the algalcoverage_Sparse and multiply by 0.300, take the biodiversityindex_High and multiply by -0.630, take the biodiversityindex_Low and multiply by 0.432, take the biodiversityindex_Medium and multiply by 0.015, take the biodiversityindex_Very High and multiply by -0.108, take the bleachingeventsperyear_Constant and multiply by 0.510, take the bleachingeventsperyear_Frequent and multiply by -0.053, take the bleachingeventsperyear_Occasional and multiply by -0.287, take the bleachingeventsperyear_Rare and multiply by -0.461, take the coralageestimate and multiply by -0.028, take the coraldominanttype_Branching and multiply by -0.638, take the coraldominanttype_Encrusting and multiply by 0.096, take the coraldominanttype_Massive and multiply by 0.120, take the coraldominanttype_Soft and multiply by 0.131, take the currentflowquality_Excellent and multiply by -0.995, take the currentflowquality_Fair and multiply by 0.163, take the currentflowquality_Good and multiply by -0.063, take the currentflowquality_Poor and multiply by 0.605, take the distantstressindicators_Confirmed and multiply by -0.399, take the distantstressindicators_None and multiply by 0.046, take the distantstressindicators_Possible and multiply by 0.062, take the invasivespeciesdetected_No and multiply by -0.861, take the invasivespeciesdetected_Yes and multiply by 0.570, take the nearbyhealthyreef_Absent and multiply by 0.214, take the nearbyhealthyreef_Adjacent and multiply by -0.923, take the nearbyhealthyreef_Nearby and multiply by 0.418, take the observationduration and multiply by 0.013, take the pollutionlevel_Critical and multiply by 0.433, take the pollutionlevel_High and multiply by -0.267, take the pollutionlevel_Low and multiply by -0.345, take the pollutionlevel_Moderate and multiply by -0.112, take the predatoractivitylevel_Catastrophic and multiply by 0.067, take the predatoractivitylevel_Extreme and multiply by -0.411, take the predatoractivitylevel_High and multiply by 0.633, take the predatoractivitylevel_Low and multiply by -0.135, take the predatoractivitylevel_Moderate and multiply by 0.111, take the predatoractivitylevel_Moderate High and multiply by 0.087, take the predatoractivitylevel_Moderate Low and multiply by -0.204, take the predatoractivitylevel_None and multiply by 0.599, take the predatoractivitylevel_Severe and multiply by -0.127, take the predatoractivitylevel_Very Low and multiply by -0.911, take the previousstressincidents_Major and multiply by -0.225, take the previousstressincidents_Minor and multiply by 0.102, take the previousstressincidents_Moderate and multiply by 0.263, take the previousstressincidents_None and multiply by -0.430, take the reefaverageage and multiply by -0.011, take the reefdepthzone_Deep and multiply by -0.460, take the reefdepthzone_Mid-depth and multiply by -0.141, take the reefdepthzone_Shallow and multiply by 0.310, take the reefintegrityscore_Excellent and multiply by -1.016, take the reefintegrityscore_Good and multiply by -0.386, take the reefintegrityscore_Moderate and multiply by 0.045, take the reefintegrityscore_Poor and multiply by 0.482, take the reefintegrityscore_Very Poor and multiply by 0.585, take the reefmonitoringduration_1-3 years and multiply by 0.201, take the reefmonitoringduration_4-6 years and multiply by 0.087, take the reefmonitoringduration_< 1 year and multiply by -0.473, take the reefmonitoringduration_> 6 years and multiply by -0.106, take the remotesensorpresent_No and multiply by 0.005, take the remotesensorpresent_Yes and multiply by -0.296, take the surveyorexperience_No and multiply by 0.048, take the surveyorexperience_Yes and multiply by -0.339, add those up and if the number is greater than -0.529 then predict Healthy, otherwise predict Stressed.

Decision Tree

Dummy

Always predict 0

RuleFit

Bayesian Rule List

CORELS

1. {'col': 'X0', 'index_col': 0, 'cutoff': 0.5, 'val': 0.12179487179487179, 'flip': True, 'val_right': 0.4140127388535032, 'num_pts': 783, 'num_pts_right': 471}
2. {'col': 'X37', 'index_col': 37, 'cutoff': 0.5, 'val': 0.05333333333333334, 'flip': True, 'val_right': 0.18518518518518517, 'num_pts': 312, 'num_pts_right': 162}
3. {'col': 'X19', 'index_col': 19, 'cutoff': 0.5, 'val': 0.0472972972972973, 'flip': False, 'val_right': 0.5, 'num_pts': 150, 'num_pts_right': 2}
4. {'col': 'X59', 'index_col': 59, 'cutoff': 0.5, 'val': 0.02654867256637168, 'flip': False, 'val_right': 0.11428571428571428, 'num_pts': 148, 'num_pts_right': 35}
5. {'col': 'X12', 'index_col': 12, 'cutoff': 0.5, 'val': 0.01818181818181818, 'flip': False, 'val_right': 0.3333333333333333, 'num_pts': 113, 'num_pts_right': 3}
6. {'val': 0.01818181818181818, 'num_pts': 110}

EBM