Методы машинного обучения в риск-профилировании инвестора

AlterEconomics (ранее - Журнал экономической теории). 2024. Т. 21. № 3. С. 527–552.

Kurochkin S. V., Павлов Н., Mariya T., Яремич Е.

Investment profiling is essential for investors as it differentiates between financial products that align with their risk tolerance and those that are excessively risky. Additionally, investment profiling serves as a tool to prevent misselling. Most financial institutions use risk profiling questionnaires to establish an investor's risk profile. However, the effectiveness of this method is questionable, as actual investor behavior can significantly differ from the responses provided in these questionnaires. This article aims to develop an interactive platform for investment profiling, where data is collected through a web interface. Regression analysis is conducted using Python. On the platform, users engage in a game simulating exchange trading, where they must select a stop loss or take profit in each of ten rounds. This approach allows the investor's risk profile to be determined based on actual user behavior rather than abstract questionnaire responses. The users' actions are then classified using machine learning methods. As a result, it was shown that the constructed model correctly predicts 65.78 % of investors’ decisions, whereas the survey correctly identified the risk profile of only 53.7 % of investors. These findings could be used by financial companies for the improvement of their investment profiling process.

Keywords: искусственный интеллект, машинное обучение, анкетирование, machine learning, толерантность к риску, questionnaire, artificial intelligence, risk profiling, risk tolerance, риск-профиль