Quantitative Trading of Stocks Based on TD3 Algorithm

Abstract

Intelligent and efficient stock analysis can help investors and institutions judge stock trends, improve investment returns, and avoid investment risks. The use of deep reinforcement learning methods to process stock data and provide investment recommendations has important research value. This question proposes to use the TD3 algorithm to implement a deep reinforcement learning model, introduce commonly used stock technical indicators, design reward and action functions, and use them to backtest recent stock trading data. Finally, comparing it with the moving average strategy and other deep reinforcement learning models, it was found that in the past decade of historical data, the annualized rate of the moving average strategy was 10% -15%, while the annualized rate of the TD3 algorithm was 23% -25%. This indicates that the TD3 algorithm can help investors or institutions make judgments and effectively improve investment returns.