Sberbank offers to predict revenue| 15.07.2019
The corporate & investment banking unit of Sberbank has developed a model for forecasting earnings of the companies, using machine learning based on transaction data. It allows you to obtain forecasts of revenue
the current year for all active was nine months before the results are officially published.
In this case, the model was used class of black box, based on decision trees — Random Forest Regression. It is built for all 8 million companies (entrepreneurs and legal entities) that are present in
the Russian market.
According to the press service of the savings Bank, such an early analysis of funds provides the opportunity to plan ahead work with the companies. Besides, with the new model it is possible to obtain forecasts including on IP,
information which is not present in any open source.
Business intelligence (BI), financial instruments
Journal: Journal IT-News, Subscription to magazines
Sberbank | Sberbank