Learning financial asset-specific trading rules via deep reinforcement learning

Generating asset-specific trading signals based on the financial conditions of assets is one challenging problems in automated trading. Various asset rules are proposed experimentally different technical analysis techniques. However, these kind strategies profitable, extracting new from vast historical data to increase total return and decrease risk portfolios difficult for human experts. Recently, various deep reinforcement learning (DRL) methods employed learn each asset. In this paper, a novel DRL model with feature extraction modules proposed. The effect input representations performance models investigated DRL-based markets situations studied. work outperformed other state-of-the-art single obtained almost 12.4% more profit over best Dow Jones Index same time period.