The Design and Implementation of Quantum Finance-based Hybrid Deep Reinforcement Learning Portfolio Investment System

Authors:
Open Access
Publisher website
Google Scholar
Abstract
With the rapid development of Deep Learning (DL) and Reinforcement Learning (RL), Deep Reinforcement Learning (DRL) has become one of the leading trends in automatic trading. However, research of RL in portfolio investment has difficulties in distributing investment funds, controlling profit and loss, and exploring the unseen environment. This paper introduces an intelligent portfolio investment system based on the integration of DRL, Quantum Finance Theory (QFT). Our proposed system consists of two agents: 1) A trading agent based on Deep Deterministic Policy Gradient (DDPG) algorithm to generate continuous actions for investment weighting; and 2) A risk-control agent based on Policy Gradient (PG) algorithm produces discrete actions according to each day’s Quantum Price Levels (QPLs). One significant merit of integrating two intelligent agents is that they can cooperate to make more reasonable and stable fund distribution adjustments in the portfolio investment. The experimental results reflect the flexibility and robustness of our system, as it achieves considerable profits in back-tests consisting of various combinations of FOREX products. (Source code is available at: https://github.com/772435284/QF-portfolio-investment-system)

This publication has 5 references indexed in Scilit: