CAREER LAUNCHPAD PROJECT : Customer Churn Prediction for Telecom Company

A comprehensive end-to-end machine learning project to predict customer churn in the telecom industry using advanced feature engineering, multiple ML models, and interactive visualizations.

Project Objectives :

• Predict Customer Churn: Identify customers likely to leave the service
• Feature Engineering: Create meaningful features including tenure groups and LTV segmentation
• Model Comparison: Train and compare Logistic Regression and Random Forest models
• Interpretability : Use SHAP values to explain model predictions
• Interactive Dashboard : Provide a user-friendly interface for predictions and insights Quick Start

Installation

1. Clone or navigate to the project directory:

cd CUSTOMER_CHURn

2. Create a virtual environment (recommended):

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

3. Install dependencies:

pip install -r requirements.txt

Usage

Option 1: Run the Complete Pipeline

python main.py --mode all

Option 2: Step-by-Step Execution

# Generate synthetic dataset
python main.py --mode generate_data

# Train models
python main.py --mode train

# Evaluate models
python main.py --mode evaluate

# Launch dashboard
python main.py --mode dashboard

Option 3: Use Jupyter Notebooks

jupyter notebook

Then open notebooks in order:

1. 01_data_exploration.ipynb
2. 02_feature_engineering.ipynb
3. 03_model_training.ipynb Option 4: Launch Dashboard Directly

streamlit run dashboard/streamlit_app.py

Project Structure

CUSTOMER_CHURn/
├── data/
│   ├── raw/                    # Raw dataset
│   └── processed/              # Processed datasets
├── notebooks/
│   ├── 01_data_exploration.ipynb
│   ├── 02_feature_engineering.ipynb
│   └── 03_model_training.ipynb
├── src/
│   ├── data_loader.py          # Data loading and generation
│   ├── preprocessing.py        # Data preprocessing
│   ├── feature_engineering.py  # Feature engineering
│   ├── model_training.py       # Model training pipeline
│   ├── model_evaluation.py     # Evaluation metrics
│   └── interpretability.py     # SHAP analysis
├── models/
│   └── saved_models/           # Trained model artifacts
├── visualizations/
│   └── plots/                  # Generated plots
├── dashboard/
│   └── streamlit_app.py        # Interactive dashboard
├── requirements.txt
├── README.md
└── main.py                     # Main execution script

Key Features:

1. Advanced Feature Engineering

• Tenure Groups: Categorize customers by contract length (0-12, 12-24, 24-48, 48+ months)
• LTV Segmentation: Calculate Customer Lifetime Value and create segments (Low, Medium, High, Premium)
• Interaction Features: Charges per service, contract-payment combinations
• Service Aggregation: Total services used, service diversity metrics

2. Machine Learning Models

• Logistic Regression: Fast, interpretable baseline model
• Random Forest: Powerful ensemble model for complex patterns
• Cross-Validation: 5-fold stratified CV for robust evaluation
• Hyperparameter Tuning: GridSearchCV for optimal parameters

3. Model Evaluation

• Metrics: Accuracy, Precision, Recall, F1-Score, ROC-AUC
• Visualizations: Confusion matrices, ROC curves, Precision-Recall curves
• Comparison: Side-by-side model performance analysis

4. Model Interpretability (SHAP)

• Summary Plots: Global feature importance
• Force Plots: Individual prediction explanations
• Dependence Plots: Feature interaction analysis
• Waterfall Plots: Detailed contribution breakdown

5. Interactive Dashboard

• Data Explorer: Filter and visualize customer data
• Model Performance: Real-time metrics and charts
• Churn Prediction: Single and batch customer predictions
• Feature Importance: Interactive SHAP visualizations
• Customer Insights: Segment-wise churn analysis

Dataset Features :

The synthetic telecom dataset includes:

• **Demographics: Gender, SeniorCitizen, Partner, Dependents
• **Services: PhoneService, InternetService, OnlineSecurity, TechSupport, etc.
• **Account: Contract type, PaymentMethod, PaperlessBilling
• **Charges: MonthlyCharges, TotalCharges
• **Tenure: Months with the company
• **Target: Churn (Yes/No)

Learning Outcomes :

This project demonstrates:

• End-to-end ML pipeline development
• Advanced feature engineering techniques
• Model training and hyperparameter tuning
• Model evaluation and comparison
• Explainable AI with SHAP
• Interactive dashboard development
• Best practices in ML project structure

Expected Results :

• Model Accuracy: 78-85% (depending on data quality)
• ROC-AUC Score: 0.80-0.88
• Key Churn Indicators: Contract type, tenure, monthly charges, internet service

Technologies Used :

• Python 3.8+
• pandas & numpy: Data manipulation
• scikit-learn: Machine learning
• SHAP: Model interpretability
• matplotlib & seaborn: Static visualizations
• Plotly: Interactive charts
• Streamlit: Web dashboard

Notes :

• The project includes a synthetic dataset generator for demonstration
• All code is beginner-friendly with detailed comments
• Notebooks provide step-by-step explanations
• Dashboard is production-ready and can be deployed

Contributing :

Feel free to enhance this project by:

• Adding more ML models (XGBoost, Neural Networks)
• Implementing additional features
• Improving the dashboard UI
• Adding automated testing

License:

This project is for educational and demonstration purposes.

---