GamuNyatsanza/ZimbabweMobileMoneyPredictionChurn

Predicting Zimbabwe mobile money customer churn with machine learning to boost retention. # Zimbabwe Mobile Money Customer Churn Prediction Predicting customer churn for **EcoCash**, **OneMoney**, and **TeleCash** using machine learning. --- ## Description This project predicts which mobile money customers in Zimbabwe are likely to churn. It leverages machine learning techniques to help providers **retain customers**, **reduce churn costs**, and **understand key churn drivers**. The system is designed to be **scalable**, **interpretable**, and **easy to integrate** with existing mobile money platforms. --- ## How It Works ### System Overview ```text Data Collection --> Data Preprocessing --> Feature Engineering --> Model Training (XGBoost) --> Prediction & SHAP Explanation --> Customer Segmentation --> Retention Dashboard Data Collection – Historical transaction and customer data from mobile money platforms. Data Preprocessing – Cleaning, handling missing values, and balancing classes using SMOTE. Feature Engineering – Generating features that capture customer behavior and engagement. Model Training – Using XGBoost to predict churn likelihood. Prediction & Explanation – Each prediction is accompanied by a SHAP explanation to highlight top churn risk factors. Customer Segmentation – Classifies customers into risk groups for targeted retention strategies. Retention Dashboard – Interactive frontend built with React to visualize churn predictions and insights. Visuals Backend Dashboard (API responses) Frontend Dashboard Replace these with your actual screenshots stored in docs/. Features Predict churn using XGBoost – accurate machine learning classification. Handles class imbalance with SMOTE – ensures minority churn cases are detected. Provides top churn risk factors per customer with SHAP – transparent model explanations. Segments customers for targeted retention campaigns – actionable insights for marketing. Interactive Dashboard – visualizes customer churn risk and insights. Tech Stack Python – scikit-learn, XGBoost, SHAP …