Оценка поворотных точек индикаторов деловой активности Банка России с применением методов машинного обучения

Деньги и кредит. 2026. Т. 85. № 2. С. 3–36.

Zvereva V., Крупкина А. С., Андреев А. В., Семитуркин О. Н., Кудаева М. С.

This study develops a methodology for identifying threshold values of Bank of Russia business activity indicators to determine business cycle phases using machine learning classification models. The study relies on monthly monitoring of businesses data from the Bank of Russia for the period from January 2009 to September 2025. The greatest contribution to the model predictions comes from the following monitoring indicators: enterprises’ assessments of actual demand for products, business climate indicators, and enterprise expectations regarding changes in production volumes over the next three months. Comparative accuracy analysis shows the systematic superiority of ensemble methods (e.g. bagging and various types of boosting) over parametric models. The obtained threshold values make it possible to formalise and strengthen the analytical basis for expert judgements on the current phase of the business cycle using Bank of Russia survey data.

Keywords: денежно-кредитная политика, экономический цикл, опросы предприятий, экономическая активность, economic activities classification, business surveys of entrepreneurs, Machine learning algorithms, Bank of Russia monetary policy, economic cycles and crisis, анализ данных и машинное обучение