The churn prediction is an important problem, which can be found in businesses with a constant stream of customers that use subscription services: banking, telecommunication and entertainment industry. The focus of business development shifts from attracting the new customers to retention of the old ones and therefore modeling users' outflow can be used to plan company's tactics and strategies. However, it is often more important not to just know the outflow indicators on a macro level, but to predict the churn probability for every customer to spot interventions. The object of this book is the process of the customer churn. The subject of research can be defined as automated models for the churn prediction and its practical use. The models were designed on a basis of state-of-the-art machine learning methods: the ensemble trees methods and the logistic regression. The Metric for the model evaluation is proposed; process of the model optimization on this metric is well-described.The main result of this work is the cohort-based ensemble model for the churn prediction and automation of the model construction pipeline.