Перспективи моделей навчання з підкріпленням у задачах підтримки прийняття інвестиційних рішень

Abstract

The investigation aims to evaluate the effectiveness of machine learning algorithms in investment decision-making by examining reinforcement learning models within the investment decision support system. The research includes an analysis of the PPO, IMPALA, and A2C models. The Gym environment was used to simulate automated trading, and it was found that PPO was the most effective in maximizing portfolio profitability. The findings from 30 experiments were compared using Welch's t-test to ensure objectivity and robustness.