📉 Customer Churn Prediction

An end-to-end machine learning project that performs EDA, compares multiple models (Logistic Regression, Random Forest, XGBoost), tunes hyperparameters, and deploys a real-time churn prediction app using Streamlit — achieving 86% ROC-AUC on the IBM Telco dataset.

🔗 Live Demo: customer-churn-prediction-pranjal-pandey-003.streamlit.app

---

📌 Project Overview

Customer churn is one of the biggest challenges in the telecom industry. This project builds a predictive ML pipeline that identifies at-risk customers based on their demographics, services, and billing information — helping businesses take proactive retention measures.

---

🚀 Features

• Predicts churn probability in real-time
• Interactive UI with dropdowns and sliders for all customer attributes
• Risk level categorization: 🔴 High / 🟡 Medium / 🟢 Low
• XGBoost model trained with class imbalance handling
• Model trains on startup — no pickle version issues

---

🛠️ Tech Stack

Layer	Tools
ML Model	XGBoost, Scikit-learn
Data Processing	Pandas, NumPy
Frontend	Streamlit
Deployment	Streamlit Cloud
Version Control	Git, GitHub

---

📊 Models Compared

Model	ROC AUC	F1 Score	Recall
Logistic Regression	0.86	0.64	0.84
Random Forest (Tuned)	0.85	0.65	0.78
XGBoost ✅	0.86	0.65	0.80

XGBoost selected as final model for best overall balance of precision, recall and AUC.

---

📂 Project Structure

Customer-Churn-Prediction/
├── app.py                        # Streamlit frontend + model training
├── requirements.txt              # Dependencies
├── Telco_Customer_Churn.csv      # Dataset (IBM Telco)
├── Customer_Churn_Prediction.ipynb  # EDA + model experiments
└── .gitignore

---

📥 Input Features

Demographics: Gender, Senior Citizen, Partner, Dependents

Services: Phone Service, Multiple Lines, Internet Service, Online Security, Online Backup, Device Protection, Tech Support, Streaming TV, Streaming Movies

Billing: Contract Type, Paperless Billing, Payment Method, Monthly Charges, Total Charges, Tenure

---

⚙️ Run Locally

# Clone the repo
git clone https://github.com/venom312004/Customer-Churn-Prediction.git
cd Customer-Churn-Prediction

# Create virtual environment
python -m venv venv
venv\Scripts\activate      # Windows
# source venv/bin/activate # Mac/Linux

# Install dependencies
pip install -r requirements.txt

# Run the app
streamlit run app.py

---

📈 Dataset

IBM Telco Customer Churn Dataset

• 7,043 customers
• 21 features
• Target: Churn (Yes/No)
• Class imbalance handled via scale_pos_weight in XGBoost

---

👤 Author

Pranjal Pandey

• GitHub: @venom312004
• B.Tech Data Science & AI | GGSIPU Delhi