Industrial Automatic Control Systems and Controllers Annotation << Back Predicting Bank Customer Churn Using Machine Learning Methods Bazhenov R.I., Velikovskaya E.V., Sedova N.A. Since attracting a new client is much more expensive than retaining existing ones, there is a pressing problem of predicting churn by analyzing preferences for specifi c services. This paper presents the results of predicting the churn of bank clients using modern machine learning methods. Such a forecast will predetermine the process of leaving the bank client, allowing the bank to take measures in advance to retain him. The training sample, containing over ten thousand depersonalized data on bank clients (taken from open sources), includes 20 identifi cation features, by which the forecast is carried out. The paper presents a description of these identifi cation features, as well as the results of training by seven machine learning methods. The paper presents assessments of the quality of forecasting the churn of bank clients for all the machine learning methods used, while computer modeling showed that the best method for the problem being solved is extreme gradient boosting. Keywords: machine learning method, customer churn, area under the curve, ensemble of methods, extreme gradient boosting, random forest. DOI: 10.25791/asu.10.2024.1535 Pp. 19-24.