Application of Deep Reinforcement Learning in Financial Quantitative Trading

Abstract

Quantitative trading refers to the use of computer technology and mathematical algorithms to carry out financial transactions. Investors formulate trading strategies based on historical data to avoid irrational investment decisions made by investors due to subjective emotional fluctuations as much as possible, and ultimately maximize profits. In recent years, artificial intelligence technology has been developing rapidly. Inspired by Alpha Go decision game theory, this paper introduces deep reinforcement learning algorithm into the field of financial quantitative trading. Deep reinforcement learning algorithm can effectively solve the learning problem with long-term goals and delayed reward, and has high applicability for financial stock market. The proposed deep reinforcement learning algorithm adopts the Deep Q Network (DQN), and the reward function is carefully designed for different output actions, so that the model can better capture the hidden dependencies and potential dynamics in stock data. The double-average trading strategy is adopted to improve the accuracy of decision-making, and ultimately maximize profits by learning the trading rules. In order to evaluate the superiority of the model, this paper compares it with other deep reinforcement learning algorithms. Finally it verifies the advantages of DQN algorithm in the decision-making environment of stock trading and the feasibility of applying deep reinforcement learning model to financial quantitative trading.