About

This canvas implements a comprehensive customer churn prediction system that trains ensemble machine learning models on telecom data, analyzes feature importance across multiple algorithms, and deploys batch predictions with RAG-powered insights. The workflow combines data cleaning, three-model ensemble training (Random Forest, Gradient Boosting, AdaBoost), rigorous evaluation metrics (cross-validation, calibration, ROC-AUC), business impact quantification, and a retrieval-augmented generation (RAG) system that delivers personalized churn risk tiers and retention recommendations for individual customers and batch processing.