Churn Prediction System

A complete customer-churn prediction project — from data preprocessing and model training, to deployment via a Streamlit web app. This project demonstrates building a full-pipeline ML solution with feature engineering, model comparison, and a user-friendly interface for real-time predictions.

Purpose

The goal of this project is to predict whether a customer is likely to churn (cancel service) using historical telecom data. By training and comparing several models, and then deploying the best one, this system can help businesses identify at-risk customers and intervene proactively — improving retention and reducing revenue loss.

Repository Structure

Churn-Prediction-System/
│
├── data/
│ └── Telco_Customer_Churn_Dataset.csv
│
├── models/
│ ├── xgb_churn_model.pkl   # Trained XGBoost model
│ ├── scaler.pkl            # Standard Scaler
│ └── feature_names.pkl     # List of features names
│
├── .gitignore              # gitignore file
├── main.ipynb              # Jupyter notebook with data cleaning, EDA, model training & evaluation
├── main.py                 # Streamlit App
└── requirements.txt        # Dependencies

Results

Model	Accuracy	Recall (Churn = 1)	Comments
Logistic Regression	~72%	81%	Good at catching churners (high recall)
Random Forest	~78%	50%	Strong overall accuracy but misses many churn cases
XGBoost (final)	~75%	78%	Balanced performance; selected as production model

The final XGBoost model was chosen because it offered a strong trade-off between precision, recall, and overall stability after cross-validation.

How to Run the Project Locally

1. Clone the repo

git clone https://github.com/YourUsername/Churn-Prediction-System.git  
cd Churn-Prediction-System

2. Create virtual environment

python -m venv venv
source venv/bin/activate   # (Windows: venv\\Scripts\\activate)

3. Install dependencies

pip install -r requirements.txt

4. Run the Streamlit app

streamlit run app.py

Then open your browser at http://localhost:8501 to access the web interface. You can input customer details (tenure, monthly charges, contract type, etc.) and get a churn prediction with probability.

How It Works

Numeric features (tenure, MonthlyCharges, TotalCharges) are standardized using the saved scaler.pkl.

Categorical features (Contract, PaymentMethod, etc.) are one-hot encoded, and the resulting feature vector is aligned via feature_names.pkl to match the model’s expected input.

The trained XGBoost model in xgb_churn_model.pkl makes the prediction — either churn (1) or no churn (0) — and returns a probability score.

Demo

Video Showcasing Deployed Streamlit App:

Churn_Prediction_video.mp4