Predicting New Customers

🔮 Predicting New Customer Churn

Once the model is trained, you can use it to predict the churn risk for new or existing customers. This allows you to identify high-risk accounts and intervene before they cancel their service.

Input Features

To generate a prediction, you must provide a customer profile containing the following features:

| Feature | Description | Example Value | | :--- | :--- | :--- | | Contract | Length of the service agreement | 'Monthly', 'One Year', 'Two Year' | | SupportCalls | Number of interactions with customer support | 4 | | MonthlyBill | Current monthly billing amount (INR) | 110.0 | | PaymentMethod | Method of payment | 'CreditCard', 'BankTransfer', 'MailedCheck' | | BillingIssues | Binary flag for historical billing disputes | 0 (No) or 1 (Yes) | | DataUsageGB | Monthly data consumption in Gigabytes | 85.0 | | TenureMonths | Number of months the customer has stayed | 12 | | AutoPay | Whether the customer uses automated payments | 1 (Yes) or 0 (No) |

Usage Example

The model expects categorical variables to be transformed using the LabelEncoder instances stored during training. Use the following pattern to pass new data to the model:

import pandas as pd

# 1. Define the customer profile (Raw Data)
raw_data = {
    'Contract': 'Monthly',
    'SupportCalls': 5,
    'MonthlyBill': 125.0,
    'PaymentMethod': 'CreditCard',
    'BillingIssues': 1,
    'DataUsageGB': 150.0,
    'TenureMonths': 3,
    'AutoPay': 0
}

# 2. Transform categorical data using the le_dict encoders
new_customer = pd.DataFrame({
    'Contract': [le_dict['Contract'].transform([raw_data['Contract']])[0]],
    'SupportCalls': [raw_data['SupportCalls']],
    'MonthlyBill': [raw_data['MonthlyBill']],
    'PaymentMethod': [le_dict['PaymentMethod'].transform([raw_data['PaymentMethod']])[0]],
    'BillingIssues': [raw_data['BillingIssues']],
    'DataUsageGB': [raw_data['DataUsageGB']],
    'TenureMonths': [raw_data['TenureMonths']],
    'AutoPay': [le_dict['AutoPay'].transform([raw_data['AutoPay']])[0]]
})

# 3. Generate Prediction
prediction = model.predict(new_customer)
result = le_dict['Churn'].inverse_transform(prediction)[0]

print(f"Prediction Result: {result}")

Interpreting the Output

• Predicted Churn: Yes: The customer exhibits behavior highly correlated with churn (e.g., high support calls, monthly contract). Immediate retention efforts are recommended.
• Predicted Churn: No: The customer is likely to stay. Continue standard engagement.

Note: For bulk predictions, you can pass a DataFrame with multiple rows to model.predict() to receive an array of churn statuses for an entire customer segment.