EJ-Odeke/xai-paygo-solar-churn-prediction

Master's capstone project demonstrating an Explainable AI (XAI) framework for predicting customer churn in Uganda's PAYGo solar sector using machine learning, feature engineering, SHAP explainability, and an interactive Streamlit dashboard. # Explainable AI for Customer Churn Prediction in Uganda's PAYGo Solar Sector ## Master's Capstone Project An end-to-end machine learning project that predicts customer churn in Uganda's PAYGo solar sector using Explainable Artificial Intelligence (XAI). The project combines data preprocessing, feature engineering, predictive modeling, SHAP explainability, and an interactive Streamlit dashboard to support data-driven customer retention strategies. --- ## Project Overview Customer churn remains one of the biggest challenges facing PAYGo solar providers. Losing customers reduces recurring revenue, increases customer acquisition costs, and limits sustainable growth. This project develops an explainable machine learning framework capable of identifying customers at risk of churn while providing transparent explanations for every prediction. The framework enables business stakeholders to understand not only **who is likely to churn**, but also **why**. --- ## Objectives - Predict customer churn using machine learning models. - Engineer meaningful behavioural and financial features. - Compare multiple classification algorithms. - Improve performance through hyperparameter tuning. - Explain model predictions using SHAP (SHapley Additive Explanations). - Deploy an interactive Streamlit dashboard for business users. --- ## Technologies Used - Python - Pandas - NumPy - Scikit-learn - XGBoost - SHAP - Matplotlib - Plotly - Streamlit - Joblib --- ## Project Structure ``` . ├── data/ │ ├── raw/ │ └── processed/ │ ├── notebooks/ │ ├── paygo-solar-ai-capstone.ipynb │ ├── models/ │ ├── best_model.pkl │ ├── scaler.pkl │ └── shap_explainer.pkl │ ├── dashboard/ │ └── app.py │ ├── images/ │ ├── requirements.txt ├── README.md └── LICENSE ``` --- ## Machine Learning Workflow 1. Data Collection 2. Data Cleaning 3. Exploratory Data Analysis 4. Feature Engineering 5. Feature Selection 6. Model Training 7. Model Evaluation 8. Hyperparameter Tuning 9. Model Exp …