📉 ChurnGuard — Customer Churn Prediction

A machine learning classification project that predicts which telecom customers are likely to churn. Compares Logistic Regression, Random Forest and Gradient Boosting on 7,043 real customer records.

---

📋 Table of Contents

• 🎯 Project Overview
• 📊 Key Questions Answered
• 📈 Visualizations
• 🛠️ Technologies Used
• 📁 Project Structure
• 🚀 How to Run
• 💡 Key Findings
• 👨‍💻 Author

---

🎯 Project Overview

This project applies binary classification to predict customer churn for a telecom company. The dataset contains 7,043 customers with features covering demographics, account information, subscribed services and billing details.

The analysis covers:

• Exploratory data analysis of churn patterns
• Data cleaning, feature engineering and scaling
• Model training — Logistic Regression, Random Forest and Gradient Boosting
• Model comparison using accuracy and 5-fold cross-validation
• Feature importance analysis to identify key churn drivers

---

📊 Key Questions Answered

• What is the overall churn rate?
• How does contract type affect churn?
• Do monthly charges and tenure influence churn behaviour?
• Which model best predicts customer churn?
• What are the most important features in predicting churn?

---

📈 Visualizations

Churn Overview

Tenure and Monthly Charges by Churn Status

Model Comparison - Accuracy vs CV Score

Confusion Matrix

Top 15 Feature Importances - Random Forest

---

🛠️ Technologies Used

• Language: Python 3.12
• Data Manipulation: Pandas, NumPy
• Machine Learning: Scikit-learn
• Visualization: Matplotlib, Seaborn
• Environment: Jupyter Notebook

---

📁 Project Structure

ChurnGuard/
├── analysis.ipynb          ← Main analysis notebook
├── requirements.txt
├── LICENSE
├── README.md
├── data/
│   └── WA_Fn-UseC_-Telco-Customer-Churn.csv   ← Raw dataset (not tracked by git)
└── outputs/
    ├── churn_overview.png
    ├── tenure_charges.png
    ├── model_comparison.png
    ├── confusion_matrix.png
    └── feature_importance.png

---

🚀 How to Run

1. Install dependencies:

pip install -r requirements.txt

2. Download the dataset:

Get the CSV from Kaggle and place it inside the data/ folder.

3. Run the notebook:

jupyter notebook analysis.ipynb

Run all cells top to bottom. Charts will be saved automatically to outputs/.

---

💡 Key Findings

• Overall churn rate is approximately 26% across all customers
• Month-to-month contracts have a significantly higher churn rate than one or two-year contracts
• Fibre optic internet service customers churn at a higher rate than DSL customers
• Customers with shorter tenure and higher monthly charges are more likely to churn
• Tenure, monthly charges and contract type are the top predictors of churn

---

👨‍💻 Author

Berke Arda Turk
Data Science & AI Enthusiast | Computer Science (B.ASc)
🌐 Portfolio · 💼 LinkedIn · 🐙 GitHub