One more successful machine learning use case in banking was with the largest Russian bank — Sberbank. Fayrix's team had to build a prediction model of the total and individual customers' spending. To solve the problem, an ensemble of models was developed: forecasting time series, linear regression, decision trees. The goals of this solution were to increase sales through relevant personal offers and increase loyalty. The essence of the implemented method was to predict the client's transactions with a plastic card.