Credit Card Churn Prediction with GridMaster

This project is a comprehensive machine learning pipeline for predicting customer churn in a credit card dataset. It showcases the full lifecycle from data cleaning and exploratory analysis to advanced model development using GridMaster, a custom-built Python package designed to optimize and compare multiple machine learning models efficiently.

🌐 View Project Page

📁 Project Structure

credit-card-churn-prediction/
├── data/                     # Input datasets
├── notebooks/                # Jupyter Notebooks
│   ├── 01_data_cleaning.ipynb
│   ├── 02_analysis.ipynb
│   └── 03_modeling_with_gridmaster.ipynb
├── .gitignore
├── LICENSE
├── README.md
└── requirements.txt

📊 Notebooks Overview

• 01_data_cleaning.ipynb → Loads and cleans the raw dataset, transforms key variables.
• 02_analysis.ipynb → Performs exploratory data analysis (EDA), generates insights and visualizations.
• 03_modeling_with_gridmaster.ipynb → Builds and optimizes models using GridMaster, including:
  • Multi-model, multi-stage coarse-to-fine hyperparameter tuning
  • Advanced visualizations of tuning curves, performance metrics, and feature importance
  • Extraction of best models and parameters for deployment
  • Summary analysis comparing classifiers (Logistic Regression, Random Forest, XGBoost) based on recall
  • Demonstration of both new-user Quickstart case and advanced use cases for GridMaster

🚀 Key Features

• Multi-model coordination: Logistic Regression, Random Forest, XGBoost, catboost,LightGBM
• Coarse-to-fine hyperparameter tuning
• High-performance recall-focused optimization
• Visual tuning diagnostics and feature importance plots
• Best model selection and summary analysis

💡 Quickstart

1. Install dependencies:

   pip install -r requirements.txt

2. Run notebooks in order (01 → 02 → 03) to reproduce the full analysis.
3. For GridMaster usage, refer to the official user manual.

📄 License

This project is licensed under the MIT License. See the LICENSE file for details.

---

✨ Author: Winston Wang
🔗 Personal Website | GitHub