About

This canvas is a comprehensive **Telco Customer Churn Prediction ML Pipeline** that builds, trains, evaluates, and deploys an XGBoost classification model to identify at-risk telecom customers. The development layer orchestrates end-to-end model development—loading and cleaning telco data, performing exploratory data analysis with correlation and categorical feature visualizations, preprocessing and splitting datasets with stratification, training a gradient-boosted classifier with early stopping, evaluating performance metrics and generating ROC/confusion matrix visualizations, creating model explanations via permutation importance and individual customer risk analysis, and building a production-ready churn prediction function with feature importance scoring. A companion API builder layer exposes the trained model as a REST endpoint that accepts customer feature JSON and returns churn probability, risk level classification, and top contributing risk factors using SHAP-based feature importance.