src/predict.py

import argparse
import pandas as pd
import json
import sys

from src.model import ChurnClassifier

def predict_single(
    age: int,
    monthly_charges: float,
    contract_length: int,
    support_calls: int,
    tech_support: str,
    model_path: str = "models/churn_model.pkl"
):
    # Load model and preprocessor
    try:
        classifier = ChurnClassifier.load(model_path)
    except FileNotFoundError:
        print(f"Error: Model not found at path {model_path}. Please run train.py first.", file=sys.stderr)
        sys.exit(1)
        
    input_df = pd.DataFrame([{
        'age': age,
        'monthly_charges': monthly_charges,
        'contract_length': contract_length,
        'support_calls': support_calls,
        'tech_support': tech_support
    }])
    
    # Preprocess
    preprocessor = classifier.preprocessor
    X_processed = preprocessor.transform(input_df)
    
    # Predict
    churn_pred = classifier.predict(X_processed)[0]
    churn_proba = classifier.predict_proba(X_processed)[0, 1]
    
    result = {
        "prediction": int(churn_pred),
        "churn_probability": float(churn_proba),
        "status": "Churn" if churn_pred == 1 else "No Churn"
    }
    return result

if __name__ == "__main__":
    parser = argparse.ArgumentParser(description="Predict customer churn probability.")
    parser.add_argument("--model_path", type=str, default="models/churn_model.pkl", help="Path to saved model")
    parser.add_argument("--age", type=int, required=True, help="Customer age")
    parser.add_argument("--monthly_charges", type=float, required=True, help="Monthly charges in USD")
    parser.add_argument("--contract_length", type=int, choices=[1, 12, 24], required=True, help="Contract length in months")
    parser.add_argument("--support_calls", type=int, required=True, help="Number of customer support calls")
    parser.add_argument("--tech_support", type=str, choices=["yes", "no"], required=True, help="Whether tech support is active")
    
    args = parser.parse_args()
    
    res = predict_single(
        age=args.age,
        monthly_charges=args.monthly_charges,
        contract_length=args.contract_length,
        support_calls=args.support_calls,
        tech_support=args.tech_support,
        model_path=args.model_path
    )
    
    print(json.dumps(res, indent=2))