Abstract: No institution wants to lose customers and would do anything to prevent it. But banks can’t do anything about customer churn unless they predict it first. Predicting customer churn will help them to reach out to such customers and offer the assistance they need. With the support of precise and early churn prediction, customer resource management, and customer experience teams can be more assertive and customer-focused. It has been reported that simply reaching out to customers early enough can prevent 11 percent of attrition. The problem remains about how to predict such behaviour. Surveys are difficult to undertake and not many are interested in answering them. Past data holds the required solution. We can predict the behaviour of present and future customers by using machine learning and data science techniques to learn from this past customer data. In this paper, we aim to conduct a comparative analysis of different churn prediction models for banking institutions. Then we would implement a hybrid approach with the considered advantages from each of the best performing models for a given cluster of the dataset, to get better results.