ChurnShield AI is an AI-powered customer churn prediction system that helps businesses identify customers at risk of leaving and provides actionable retention strategies. Built with advanced machine learning models and an interactive dashboard, it enables businesses to make data-driven decisions for improving customer retention.
- Multi-model ensemble (XGBoost + Random Forest + Logistic Regression)
- SHAP explainability for transparent risk scoring
- Real-time probability adjustments
- Customer segmentation by risk profile
- Lifetime value preservation estimates
- Retention ROI calculator
- Personalized intervention recommendations
- Automated campaign triggers (email/SMS templates included)
- Win-back incentive simulator
- Feedback loop integration
- Model auto-refresh with new data
- Performance monitoring alerts
| Model | Accuracy | Precision | Recall | AUC |
|---|---|---|---|---|
| XGBoost | 85.5% | 83.0% | 88% | 0.85 |
| Logistic Regression | 91.7% | 80.2% | 85.3% | 0.91 |
| Random Forest | 81.1% | 82.3% | 89.9% | 0.81 |
Core ML:
- Python 3.10
- XGBoost 1.7
- Scikit-learn 1.2
- Imbalanced-learn 0.10
Dashboard:
- Streamlit 1.22
- Plotly 5.13
Original dataset enriched with:
- Customer lifetime value estimates
- Service usage trends (3-month rolling)
- Sentiment analysis from support tickets
- Network quality metrics
df['TenureToChargeRatio'] = df['tenure'] / (df['MonthlyCharges'] + 1e-6) # Avoid division by zero
df['TotalValueScore'] = (df['tenure'] * df['MonthlyCharges']) / df['TotalCharges']
df['ServiceDensity'] = df[['OnlineSecurity_Yes', 'OnlineBackup_Yes',
'DeviceProtection_Yes', 'TechSupport_Yes']].sum(axis=1) / df['tenure']
df['PaymentRisk'] = df['PaymentMethod_Electronic check'].astype(int) * df['Contract_Month-to-month'].astype(int)
df['HighCostLongTenure'] = ((df['MonthlyCharges'] > df['MonthlyCharges'].quantile(0.75)) &
(df['tenure'] > df['tenure'].median())).astype(int)# Clone repository
git clone https://github.com/codewithshami/ChurnShield_AI.git
# Set up environment
pip install -r requirements.txt
# Launch dashboard
streamlit run app.pyAccess at: http://localhost:8501
graph TD
A[New Feature] --> B{Impact Analysis}
B -->|High Value| C[Development]
B -->|Medium| D[Backlog]
C --> E[Testing]
E --> F[Deployment]
Lead Developer: Mohd Shami
Community Support:
GitHub Discussions