Abstract:This paper proposes a Deep Reinforcement Learning algorithm for financial portfolio trading based on Deep Q-learning. The algorithm is capable of trading high-dimensional portfolios from cross-sectional datasets of any size which may include data gaps and non-unique history lengths in the assets. We sequentially set up environments by sampling one asset for each environment while rewarding investments with the resulting asset's return and cash reservation with the average return of the set of assets. This enforces the agent to strategically assign capital to assets that it predicts to perform above-average. We apply our methodology in an out-of-sample analysis to 48 US stock portfolio setups, varying in the number of stocks from ten up to 500 stocks, in the selection criteria and in the level of transaction costs. The algorithm on average outperforms all considered passive and active benchmark investment strategies by a large margin using only one hyperparameter setup for all portfolios.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to apply deep reinforcement learning algorithms in high - dimensional stock portfolio trading to achieve better investment returns. Specifically, the author proposes an algorithm based on Deep Q - learning, aiming to overcome the problems of insufficient flexibility and generalization ability in existing methods when dealing with single assets or asset portfolios. The paper focuses on how to optimize investment decisions of financial assets through deep reinforcement learning, especially in the face of a large number of assets, missing data and inconsistent historical lengths, being able to effectively train the model and make it perform well on out - of - sample data. The main contributions of the paper are as follows: 1. **Propose a new deep reinforcement learning algorithm**: This algorithm can use cross - sectional data sets for effective training when dealing with high - dimensional stock portfolios, and at the same time solve the problems of possible data missing and inconsistent historical lengths in the data set. 2. **Design a unique reward mechanism**: Through the reward mechanism, the agent is encouraged to invest when predicting that an asset will perform better than the average level, and choose to hold cash when the prediction is not good, thus promoting the agent to make more favorable investment decisions. 3. **Verify the effectiveness of the algorithm**: Through out - of - sample analysis of 48 US stock portfolios of different scales, it is proved that the proposed algorithm can significantly outperform passive and active benchmark investment strategies under different numbers of stocks, selection criteria and trading cost levels. In general, this paper aims to explore how to use deep reinforcement learning techniques to improve the effectiveness of high - dimensional stock portfolio management, especially in the complex and changeable financial market environment, providing a new solution.

High-Dimensional Stock Portfolio Trading with Deep Reinforcement Learning

Deep Deterministic Portfolio Optimization

A Deep Reinforcement Learning Framework For Financial Portfolio Management

Deep Stock Trading: A Hierarchical Reinforcement Learning Framework for Portfolio Optimization and Order Execution

Evaluation of Deep Reinforcement Learning Based Stock Trading.

Practical Deep Reinforcement Learning Approach for Stock Trading

A Deep Reinforcement Learning Framework for the Financial Portfolio Management Problem

A Deep Reinforcement Learning Approach for Portfolio Management in Non‐Short‐Selling Market

Deep reinforcement learning on a multi-asset environment for trading

Deep Reinforcement Learning for Stock Portfolio Optimization

Portfolio Management using Deep Reinforcement Learning

Application of Deep Q-Network in Portfolio Management

Deep reinforcement learning for portfolio management of markets with a dynamic number of assets

Deep Reinforcement Learning for Asset Allocation in US Equities

Model-based Deep Reinforcement Learning for Dynamic Portfolio Optimization

Deep Reinforcement Learning Approach for Trading Automation in The Stock Market

A Novel Anti-Risk Method for Portfolio Trading Using Deep Reinforcement Learning

Deep Reinforcement Learning Based End-to-End Stock Trading Strategy

Online Optimal Investment Portfolio Model Based on Deep Reinforcement Learning

An Application of Deep Reinforcement Learning to Algorithmic Trading

Adaptive stock trading strategies with deep reinforcement learning methods