← Back to portfolio
Customer Churn Prediction
End-to-end churn prediction pipeline with ensemble models and class-imbalance handling.
Problem
Losing customers is expensive, and by the time churn is obvious it's often too late to act. Predicting who's at risk early gives a business room to intervene.
Approach
Built an end-to-end ML pipeline using ensemble methods (Random Forest, XGBoost), ran feature importance analysis to surface the strongest churn signals, applied SMOTE to correct for class imbalance, and tuned hyperparameters with GridSearchCV.
Results
- Ensemble comparison across Random Forest and XGBoost
- Class imbalance corrected via SMOTE before training
- Hyperparameters tuned systematically with GridSearchCV
Want the full picture, or think this kind of work fits a role you're hiring for?