Telco Customer Churn Prediction Model
ML model untuk prediksi customer churn dengan sistem retraining otomatis.
Current Version: v_initial
| Metric | Value |
|---|---|
| F1 | 0.6291 |
| Recall | 0.7166 |
| Precision | 0.5607 |
| ROC-AUC | 0.8419 |
| Threshold | 0.60 |
Structure
current/— active model artifacts (model.pkl, preprocessor.pkl, metrics.json)synthetic/— synthetic data batches for retraining experiments (SDV-generated)registry.json— version registry with history
Training
- Algorithm: LogisticRegression (sklearn 1.5.2)
- Features: 52 (after one-hot encoding)
- Train samples: 5634
- Holdout samples: 1409 (fixed, never touched)
License
MIT