| """Inference script for price increase churn prediction.""" |
| import pandas as pd |
| import numpy as np |
| import joblib |
| from features import SubscriptionFeatureEngineer |
|
|
|
|
| def predict_churn_risk(customer_data: pd.DataFrame, |
| model_path: str = 'price_increase_churn_model.pkl', |
| price_increase_pct: float = 0.15) -> pd.DataFrame: |
| """ |
| Predict cancellation risk for customers after a price increase. |
| |
| Args: |
| customer_data: DataFrame with columns matching your subscription data |
| model_path: Path to saved model |
| price_increase_pct: Percentage price increase (e.g., 0.15 for 15%) |
| |
| Returns: |
| DataFrame with churn probability and risk tier |
| """ |
| |
| engineer = SubscriptionFeatureEngineer(price_increase_pct=price_increase_pct) |
| df = engineer.transform(customer_data.copy()) |
| |
| |
| pipeline = joblib.load(model_path) |
| |
| |
| churn_prob = pipeline.predict_proba(df)[:, 1] |
| |
| |
| def risk_tier(p): |
| if p >= 0.7: |
| return 'High Risk' |
| elif p >= 0.4: |
| return 'Medium Risk' |
| else: |
| return 'Low Risk' |
| |
| results = pd.DataFrame({ |
| 'customer_id': customer_data.get('Customer ID', customer_data.index), |
| 'churn_probability': churn_prob, |
| 'risk_tier': [risk_tier(p) for p in churn_prob], |
| 'predicted_cancel_90d': churn_prob >= 0.5 |
| }) |
| |
| return results.sort_values('churn_probability', ascending=False) |
|
|
|
|
| def batch_predict_example(): |
| """Example usage with sample data.""" |
| sample = pd.DataFrame({ |
| 'Customer ID': ['C001', 'C002', 'C003'], |
| 'Tenure in Months': [3, 36, 72], |
| 'Contract': ['Month-to-Month', 'One Year', 'Two Year'], |
| 'Monthly Charge': [95.0, 65.0, 45.0], |
| 'Total Charges': [285.0, 2340.0, 3240.0], |
| 'Total Revenue': [300.0, 2500.0, 3500.0], |
| 'CLTV': [2000, 4500, 6000], |
| 'Age': [25, 45, 55], |
| 'Satisfaction Score': [1, 3, 5], |
| 'Churn Score': [85, 50, 20], |
| 'Senior Citizen': [0, 0, 1], |
| 'Dependents': [0, 1, 0], |
| 'Married': [0, 1, 1], |
| 'Partner': [0, 1, 1], |
| 'Phone Service': [1, 1, 1], |
| 'Internet Service': [1, 1, 0], |
| 'Online Security': [0, 1, 1], |
| 'Online Backup': [0, 1, 0], |
| 'Device Protection Plan': [0, 1, 1], |
| 'Premium Tech Support': [0, 0, 1], |
| 'Streaming TV': [1, 0, 0], |
| 'Streaming Movies': [1, 0, 0], |
| 'Streaming Music': [0, 0, 0], |
| 'Multiple Lines': [1, 0, 0], |
| 'Unlimited Data': [1, 0, 0], |
| 'Number of Dependents': [0, 2, 0], |
| 'Number of Referrals': [0, 2, 5], |
| 'Avg Monthly GB Download': [50, 20, 5], |
| 'Avg Monthly Long Distance Charges': [30.0, 10.0, 5.0], |
| 'Total Long Distance Charges': [90.0, 360.0, 360.0], |
| 'Total Extra Data Charges': [10, 0, 0], |
| 'Total Refunds': [0.0, 0.0, 0.0], |
| 'Population': [50000, 30000, 10000], |
| 'Quarter': ['Q3', 'Q3', 'Q3'], |
| 'Offer': ['None', 'Offer A', 'None'], |
| 'Internet Type': ['Fiber Optic', 'DSL', 'None'], |
| 'Gender': ['Male', 'Female', 'Male'], |
| 'Payment Method': ['Bank Withdrawal', 'Credit Card', 'Bank Withdrawal'], |
| 'Paperless Billing': [1, 0, 1], |
| 'Under 30': [1, 0, 0], |
| 'Referred a Friend': [0, 1, 1], |
| 'City': ['CityA', 'CityB', 'CityC'], |
| 'State': ['CA', 'NY', 'TX'], |
| 'Zip Code': ['90210', '10001', '77001'], |
| 'Country': ['United States'] * 3, |
| 'Lat Long': ['0,0', '0,0', '0,0'], |
| 'Latitude': [0.0, 0.0, 0.0], |
| 'Longitude': [0.0, 0.0, 0.0] |
| }) |
| |
| results = predict_churn_risk(sample, |
| model_path='/app/price_increase_model/price_increase_churn_model.pkl', |
| price_increase_pct=0.15) |
| print("\nSample Predictions:") |
| print(results.to_string(index=False)) |
| return results |
|
|
|
|
| if __name__ == '__main__': |
| batch_predict_example() |
|
|
