ML Pipeline for User Behavior Prediction

Developed an end-to-end machine learning pipeline leveraging XGBoost and SMOTE to accurately predict customer repurchase behavior within seven days. The pipeline includes comprehensive exploratory data analysis (EDA), detailed feature engineering, and robust experiment tracking using MLflow. Applied SHAP values for explainability, visualizing critical features like customer session details and geographical attributes. Also, addressed and resolved model pipeline errors, ensuring smooth production deployment and reproducibility through clear documentation and structured environments.