https://github.com/pydevcasts/churn_modeling_article

customer churn prediction system for banking institutions using advanced feature engineering and ensemble learning techniques. The model addresses highly imbalanced datasets (10:1 ratio) by combining SMOTE oversampling with a Soft Voting Classifier (Random Forest, Gradient Boosting, and XGBoost)
https://github.com/pydevcasts/churn_modeling_article

customer-churn-prediction ensemble-learning imbalanced-data smote

Last synced: 29 days ago
JSON representation

customer churn prediction system for banking institutions using advanced feature engineering and ensemble learning techniques. The model addresses highly imbalanced datasets (10:1 ratio) by combining SMOTE oversampling with a Soft Voting Classifier (Random Forest, Gradient Boosting, and XGBoost)

• Host: GitHub
• URL: https://github.com/pydevcasts/churn_modeling_article
• Owner: pydevcasts
• Created: 2025-06-27T13:28:37.000Z (about 1 year ago)
• Default Branch: master
• Last Pushed: 2026-05-26T23:00:43.000Z (about 1 month ago)
• Last Synced: 2026-05-27T00:20:31.905Z (about 1 month ago)
• Topics: customer-churn-prediction, ensemble-learning, imbalanced-data, smote
• Language: Jupyter Notebook
• Homepage:
• Size: 41.5 MB
• Stars: 1
• Watchers: 0
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# Customer Churn Prediction in Banking

## Advanced Feature Engineering & Ensemble Learning for Imbalanced Banking Data

[![Python 3.8+](https://img.shields.io/badge/Python-3.8+-blue.svg)](https://www.python.org/)

[![scikit-learn](https://img.shields.io/badge/scikit--learn-1.0+-orange.svg)](https://scikit-learn.org/)

[![XGBoost](https://img.shields.io/badge/XGBoost-1.5+-red.svg)](https://xgboost.ai/)

[![License](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE)

---

## 📋 Overview

This project presents a robust machine learning solution for **customer churn prediction** in the banking sector. The model addresses the challenge of highly imbalanced datasets (10:1 ratio) through advanced feature engineering and ensemble learning techniques.

### Key Highlights

- **Accuracy**: 92%

- **AUC Score**: 0.96

- **Precision**: 0.96

- **Recall**: 0.87

- **Outperformed previous best model** (91% accuracy)

---

## 🎯 Problem Statement

Customer churn is a critical challenge in the banking industry. Identifying customers likely to leave enables proactive retention strategies. The main challenges addressed:

- **Highly imbalanced dataset** (90% non-churn vs 10% churn)

- **Complex feature interactions** affecting customer behavior

- **Need for interpretable** yet powerful predictions

---

## 📊 Dataset

The dataset contains banking customer information with the following features:

| Feature | Description |

|---------|-------------|

| CreditScore | Customer's credit score |

| Geography | Country (France, Germany, Spain) |

| Gender | Male/Female |

| Age | Customer age |

| Tenure | Years with the bank |

| Balance | Account balance |

| NumOfProducts | Number of bank products used |

| HasCrCard | Credit card ownership (0/1) |

| IsActiveMember | Active member status (0/1) |

| EstimatedSalary | Estimated annual salary |

| Exited | **Target**: Churn status (1 = exited) |

---

## 🛠️ Methodology

### 1. Data Preprocessing

**Outlier Removal:**

```python

- CreditScore ≤ 359

- Age ≥ 71 years

- NumOfProducts ≥ 4