This workflow implements a complete machine learning pipeline for customer churn risk prediction following the CRISP-DM methodology. Starting from raw customer data, the pipeline performs data cleaning, feature engineering, and one-hot encoding before partitioning the dataset into a 70% training set and 30% test set using stratified sampling. Three tree-based classification models are trained and evaluated: Decision Tree, Random Forest, and XGBoost. Each model incorporates a Parameter Optimization Loop to systematically identify the best hyperparameter configuration before final evaluation on the held-out test set. Model performance is assessed using Accuracy, Macro F1, and Weighted F1, with results consolidated into an Algorithm Score Table for direct comparison across all three models.
