About

End-to-end ML pipeline predicting long-term user success
from 409,287 product analytics events.

Results:
- Gradient Boosting — AUC = 0.9964 F1 = 0.9505
- 4,771 users analysed | 22 behavioral features
- Key finding: Success users trigger 62x more credit
events than casual users
- Data leakage detected and fixed (fake AUC=1.0 → honest 0.9964)

Pipeline includes:
✓ Data cleaning ✓ Feature engineering ✓ Leakage removal
✓ 5 models trained ✓ Hyperparameter tuning
✓ 12 professional charts ✓ Business recommendations