About

Customer churn is a major challenge for subscription-based and service-oriented businesses, as losing existing customers directly impacts revenue and growth. This project focuses on building a machine learning–based customer churn prediction system that identifies customers who are likely to leave the business in advance.

The system is trained on historical customer data, including usage patterns, contract details, and billing information. A supervised classification model is used to learn churn behavior, and its performance is evaluated using validation metrics such as accuracy, recall, and ROC-AUC to ensure reliability and effectiveness.

To move beyond analysis, the trained model is published as a live API using Zerve. The API accepts customer details as input and returns a churn probability along with a clear churn risk classification, making the solution production-ready and easy to integrate with real business systems like CRM or customer support platforms.

This project demonstrates how machine learning can be transformed into a real, deployable analytics system that enables proactive customer retention strategies and data-driven decision-making.