🔮 Universal Customer Churn Predictor

Zero configuration ML pipeline - works with ANY churn dataset automatically!

✨ Key Features

• ✅ Dynamic data loading - any CSV with "Churn" target column
• ✅ Auto feature detection - sliders, dropdowns, Yes/No detected automatically
• ✅ Multi-model training - Logistic Regression, Random Forest, Gradient Boosting
• ✅ Streamlit demo - live predictions with feature importance
• Verified: Telco (11 features), Bank (9+ features), Orange Telecom (19 features)

📁 Project Structure

D:\MyProjects\Customer-Churn-Analysis\
├── app\
│   └── app.py              # Main Streamlit predictor
├── src\
│   ├── dataset_selector.py
│   ├── preprocessor.py
│   └── trainer.py
├── data\
│   └── raw\                # Drop your CSV files here
├── models\                 # Auto-generated models (.gitignore)
├── venv\                   # Virtual environment (.gitignore)
├── .gitignore              # venv, models, pycache excluded
├── requirements.txt
└── README.md

🚀 Quick Start

# 1. Activate virtual environment
D:\MyProjects\Customer-Churn-Analysis\venv\Scripts\activate

# 2. Install dependencies
pip install -r requirements.txt

# 3. Add CSV files to data/raw/
# Your datasets: customer_churn_hf, orange-telecom-20, etc.

# 4. Launch predictor
streamlit run app/app.py

🎯 How It Works

1. Select dataset → Auto-trains best model (highest AUC)
2. Smart input forms → Sliders for tenure/age, dropdowns for categorical
3. Predict → Real-time probability + business actions
4. Insights → Top risk factors + feature importance chart

✅ Verified Results

Dataset	Features	Sample Prediction	Status
customer_churn_hf	11	49-52% Medium Risk	✅
orange-telecom-20	19	96.4% HIGH RISK	✅
Bank dataset	9+	50.2% Low Risk	✅

💡 Sample High-Risk Tests

Orange Telecom (96.4% achieved):

International plan: Yes
Customer service calls: 5

Expected High Risk Patterns:

Telco: tenure=1 + Fiber optic + SeniorCitizen=Yes
Bank: IsActiveMember=No + Age=25 + Spain
Orange: International plan=Yes + Customer service calls=5+

🔧 Git Clean Configuration

✅ venv/ excluded
✅ models/*.pkl excluded  
✅ data/processed/ excluded
✅ __pycache__/ excluded
✅ .ipynb_checkpoints/ excluded
✅ .DS_Store excluded

🎉 Production Ready Features

• Universal: Works with 9, 11, 19+ features automatically
• Smart UI: Auto-detects field types (sliders/dropdowns/numeric)
• Business Actions: Emergency retention workflows
• Model Insights: Feature importance + risk explanations
• Zero Config: Drop CSV → instant predictor

Drop any churn CSV in data/raw/ → production-ready predictor in 30 seconds! 🚀