End-to-end machine learning project predicting customer churn with 84.3% ROC-AUC.
| Category | Tools |
|---|---|
| Data Prep | Excel (cleaning, encoding, one-hot) |
| EDA | Excel (pivot tables, charts) |
| ML Models | Python, scikit-learn |
| Algorithms | Logistic Regression, Decision Tree, Random Forest, Gradient Boosting |
| Evaluation | ROC-AUC, F1-Score, Confusion Matrix, Cross-Validation |
| Visualization | Matplotlib, Seaborn |
| Version Control | Git, GitHub |
| Metric | Score |
|---|---|
| Accuracy | 76.2% |
| Recall | 77.8% |
| F1-Score | 63.5% |
| ROC-AUC | 84.3% |
Best Model: Random Forest
- Month-to-month contracts = highest churn risk (17.1% importance)
- Short tenure = second strongest predictor (12.6%)
- Tech support = significant retention factor (6.9%)
├── data/ # Dataset ├── notebooks/ # Jupyter notebook ├── models/ # Saved model (.pkl) ├── reports/ # Results & visuals ├── Churn_Analysis_Report.pdf └── requirements.txt # Dependencies
pip install -r requirements.txt
jupyter notebook