Gradient Boost Comprehensive¶

Note on Results

These results are based on local validation sets provided during the competition phase and do not represent final official leaderboard standings.

Achieved the highest Dataset A AUC (0.7930) in local validation but showed significant overfitting with low cross-dataset stability (82.4%).

Cross-Dataset Performance¶

Metric	Dataset A	Dataset B
ROC AUC	0.7930	0.6533
F1 Score	0.4186	0.3721

Generalization Metric	Value
Robust Score	0.538 (Rank #13)
Stability Score	82.4%
Min AUC	0.6533
AUC Drop	-17.6%

Generalization Analysis¶

This model showed significant performance variance between datasets:

Dataset A AUC: 0.7930 (Rank #1)
Dataset B AUC: 0.6533 (Rank #6)
Dropped 5 ranks between datasets
Stability of 82.4% indicates overfitting to Dataset A characteristics

Architecture¶

flowchart TD
    A["Input Features (100+)"] --> B1["Gradient Boosting 40%"]
    A --> B2["Random Forest 35%"]
    A --> B3["Logistic Regression 25%"]

    B1 --> C["Weighted Average"]
    B2 --> C
    B3 --> C

    C --> D["Final Prediction"]

Ensemble Weights¶

Model	Weight	Role
GradientBoostingClassifier	40%	Primary predictor
RandomForestClassifier	35%	Diversity, bagging
LogisticRegression	25%	Linear baseline, calibration

Key Features¶

Comprehensive feature set including:

Segment statistics: Mean, std, skewness, kurtosis differences
Distribution tests: KS statistic, t-test, Mann-Whitney
Effect sizes: Cohen's d, Glass's delta
Stability measures: Variance ratio, IQR differences

Component Configurations¶

Gradient Boosting¶

GradientBoostingClassifier(
    n_estimators=200,
    learning_rate=0.1,
    max_depth=6,
    subsample=0.8,
    random_state=42
)

Random Forest¶

RandomForestClassifier(
    n_estimators=200,
    max_depth=10,
    min_samples_split=5,
    random_state=42
)

Logistic Regression (Calibrated)¶

CalibratedClassifierCV(
    LogisticRegression(C=0.1, max_iter=1000),
    cv=3
)

Usage¶

cd gradient_boost_comprehensive
python main.py --mode train --data-dir /path/to/data --model-path ./model.joblib

Comparison with Other Models¶

Model	Robust Score	Stability	Dataset A	Dataset B
xgb_tuned_regularization	0.715	96.3%	0.7423	0.7705
weighted_dynamic_ensemble	0.664	98.4%	0.6742	0.6849
gradient_boost_comprehensive	0.538	82.4%	0.7930	0.6533

Note: Despite having the highest Dataset A AUC, this model's low stability resulted in a robust score of only 0.538, ranking #13 out of 25 models.