Customer Churn Prediction is a task that analyses or identifies the customers who are likely to stop using the product,service,or subscription in the near future using data analytics and machine learning techniques. By analyzing customer behavior, transaction history, and engagement metrics, businesses can proactively intervene to retain at-risk customers, thereby reducing revenue loss and improving customer lifetime value. This project focuses on multi-domain churn prediction using transfer learning, which enables knowledge transfer from data-rich domains to data-scarce domains to improve the prediction accuracy. This study aims to improve churn prediction accuracy across multiple domains using transfer learning. For example, telecom dataset(data-rich) and data-scarce domains may be like new streaming services or new bank services. In this approach, the model is pre- trained on a known large labeled source dataset to learn customer behavior patterns. This model then fine-tuned on a smaller target domain dataset, optimizing for specific churn indicators in the new domain. The results demonstrate that the it shows better prediction accuracy and generalization compared to the models trained on the target dataset, the transfer learning improves churn prediction performance in data-scarce by learning the knowledge from the data-rich domains. By enabling accurate churn prediction across different industries and also assisting businesses in reducing customer churn and improving long-term customer retention.