Stock trading strategies play a critical role in investment. However, it is challenging to design a profitable strategy in a complex and dynamic stock market. In this paper, we propose a deep ensemble reinforcement learning scheme that automatically learns a stock trading strategy by maximizing investment return. We train a deep reinforcement learning agent and obtain an ensemble trading strategy using the three actor-critic based algorithms: Proximal Policy Optimization (PPO), Advantage Actor Critic (A2C), and Deep Deterministic Policy Gradient (DDPG). The ensemble strategy inherits and integrates the best features of the three algorithms, thereby robustly adjusting to different market conditions. In order to avoid the large memory consumption in training networks with continuous action space, we employ a load-on-demand approach for processing very large data. We test our algorithms on the 30 Dow Jones stocks which have adequate liquidity. The performance of the trading agent with different reinforcement learning algorithms is evaluated and compared with both the Dow Jones Industrial Average index and the traditional min-variance portfolio allocation strategy. The proposed deep ensemble scheme is shown to outperform the three individual algorithms and the two baselines in terms of the risk-adjusted return measured by the Sharpe ratio.

Stock investment can be one of the ways to manage one’s asset. Technical analysis is sometimes used in financial markets to assist traders make buying and selling decisions [1]. Many technical analysis trading rules are deterministic trading policies. [2] uses genetic algorithm to find technical trading rule. [3] studies evolutionary algorithms in optimization of technical rules for automated stock trading. [4] proposed a stock trading system based on optimized technical analysis parameters for creating buy-sell points using genetic algorithms. [5] studies the selection of the optimal trading model for stock investment in different industries. [6] describes the optimization of trading strategies. The optimization of trading rule using genetic algorithm or evolutionary algorithms belongs to policy-based method, which is a branch of Machine Learning. Policy-based methods try to directly optimize for the optimal policy which is an important branch for domains with continuous action spaces [7]. There are studies focus on how to find a trading strategy via Reinforcement Learning (RL) [8] or using Deep-Q learning for automatic trading algorithm [9]. But in this study we will focus on the policy-based method using Generic Algorithm that directly search for the optimal parameters of a deterministic policy. Yahoo Finance’s stock history data [10] will be used in this study. The reason to choose Yahoo Finance data is because it is free and available for public to assess. The performance of algorithm will be evaluated using different stocks. The purpose of the study is to see the difference between using an agent with optimized policy to manage one’s asset with buy-and-hold strategy, or manage one’s asset with an agent with unoptimized policy.

本项目复现了 2020 年 ACM 国际金融人工智能会议（ICAIF 2020）上发表的论文《Deep Reinforcement Learning for Automated Stock Trading: An Ensemble Strategy》，使用 PPO、A2C、DDPG 等多种深度强化学习算法并做集成，在道琼斯 30 只股票上进行交易决策研究，并评估其夏普比率等风险调整收益表现。

Avoy Datta, Raul Girbal, Kyle D'Souza: Optimizing Trading Ensembles: Forked from Yang et al.

Deep Reinforcement Learning for Automated Stock Trading Strategy

No description available

No description available

No description available

No description available

No description available

搬运自论文https://www.arxiv.org/pdf/2511.12120 ，仅用于论文复现

Deep Reinforcement Learning for Automated Stock Trading: An Ensemble Strategy

Playing with https://github.com/AI4Finance-Foundation/Deep-Reinforcement-Learning-for-Automated-Stock-Trading-Ensemble-Strategy-ICAIF-2020 for crypto

An implementation of the "Deep Reinforcement Learning for Automated Stock Trading: An Ensemble Strategy" paper. Uses PPO, A2C, and DDPG agents with a rolling-window validation method to optimize portfolio returns.

Deep Reinforcement Learning Ensemble for Automated Stock Trading This project implements an ensemble trading strategy using five DRL agents — PPO, DDPG, TD3, SAC, and A2C — to trade the Dow 30 constituents. The system trains each agent on in-sample data, selects the one with the highest Sharpe ratio, and deploys it on out-of-sample data.