Description
Build a complete churn prediction pipeline from feature engineering to model deployment and retention strategy.
You are a data scientist specializing in customer analytics and predictive modeling. Design a churn prediction system for: [BUSINESS TYPE: SaaS / E-commerce / Subscription / Telecom] [AVAILABLE DATA: list your data sources and key fields] [CURRENT CHURN RATE: ___] [CUSTOMER BASE SIZE: ___] Provide: 1. **Feature Engineering**: Identify the top 20 predictive features organized by category (behavioral, transactional, demographic, engagement). Include feature calculation formulas and time windows. 2. **Model Selection**: Compare at least 3 approaches (Logistic Regression, Random Forest, XGBoost, Neural Network) with pros/cons for this use case. Recommend the best approach with justification. 3. **Python Implementation**: Provide scikit-learn/XGBoost code for the complete pipeline including data preprocessing, feature scaling, model training, hyperparameter tuning with cross-validation, and evaluation metrics (AUC-ROC, precision-recall, lift charts). 4. **Interpretability**: Use SHAP values to explain predictions. Show how to generate individual customer churn risk reports. 5. **Retention Strategy**: Based on churn risk segments, recommend targeted interventions with expected ROI calculations. All code should be production-ready with proper error handling and logging.
No comments yet. Be the first!