Using Machine Learning Technologies for Predicting Customer Churn in Music Services

Currently, the study of machine learning techniques is of interest for solving the problems of various business sectors related to the processing of constantly generated and accumulated large amounts of data. In the field of music streaming, one of the most important tasks is to forecast customer churn behavior. Timely prediction of churning customers allows to take preventive measures to retain risky customers, i.e. to reduce the number of people who stop using the service (cancel the subscription). The purpose of this work is to develop a model, based on historical data on users of the service and their behavior, using machine learning techniques. The model allows to predict the probability of a person leaving the music service. The concept of a streaming service as well as the customer churn problem are studied. Review of algorithms for solving the problem of binary classification is carried out, corresponding technological tools are considered. An experimental study on a test dataset was conducted in a separate Chapter. The work may be of interest both for companies that solve the problem of customer churn and for researchers engaged in the field of machine learning. The work contains 55 pages, 3 chapters, 36 figures, 2 tables, 40 sources, 3 annexes. Keywords: Python; machine learning; churn prediction; classification; streaming service