Выявление манипуляций на российском фондовом рынке с использованием методов искусственного интеллекта

Финансы и бизнес. 2024. Т. 20. № 4. С. 57–78.

Khaziev G.

This paper examines the the eff ectiveness of artifi cial intelligence models to detect manipulation of Russian stocks based on the authors’ sample of 866 manipulation cases over the period from 2012 to 2024. We build four artifi cial intelligence classifi cation models to fi nd the most eff ective method for manipulation detection: Logistic Regression, CatBoostClassifi er, CatBoostCustom and Stacking model. Machine learning and artifi cial intelligence are eff ective in detecting manipulation in the Russian stock market (detecting over 81% of market manipulation). Models using a combination of machine learning algorithms (stacking models) demonstrate superior performance in detecting manipulation cases compared to standard machine learning models. Incorporating the intensity of stock discussions on social media into stock manipulation detection models can improve their reliability by 0,1–9,8%

Keywords: фондовый рынок, машинное обучение, stocks, stock market, манипулирование рынком, market manipulation, Artificial Intelligence, machine learning