Intelligent Portfolio Management Based on Deep Deterministic Policy Gradient
Journal of Business and Economic Research · 27.05.2026 · 2 (2)
DOI: 10.63887/jber.2026.2.2.30 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Business and Economic Research, online erschienen am 27.05.2026. Verfasst von Tang He. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).
Abstract (Originalsprache)
Portfolio management aims to maximize long-term risk-adjusted return under risk constraints. Traditional methods rely heavily on statistical assumptions and expert rules and often struggle with non-stationarity, noise, and cross-asset dependence. This paper proposes an intelligent portfolio management method based on Deep Deterministic Policy Gradient (DDPG). We formulate asset allocation as a Markov decision process, build a state space from price features, technical indicators, portfolio status, and temporal information, and design a reward that accounts for return, transaction cost, and concentration. Experiments on four U.S. stocks-Apple (AAPL), Alphabet (GOOGL), Amazon (AMZN), and IBM-over about five years show a total return of 115.33%, an annualized return of 17.19%, a Sharpe ratio of 1.044, and a maximum drawdown of 21.33%. The results show that deep reinforcement learning can solve dynamic asset allocation in continuous action spaces and has practical value for robo-advisory and quantitative investment.
Zitieren
APA
He, T. (2026). Intelligent Portfolio Management Based on Deep Deterministic Policy Gradient. Journal of Business and Economic Research, 2(2). https://doi.org/10.63887/jber.2026.2.2.30
Vancouver
He T. Intelligent Portfolio Management Based on Deep Deterministic Policy Gradient. Journal of Business and Economic Research. 2026;2(2). doi:10.63887/jber.2026.2.2.30
BibTeX
@article{He2026,
author = {He, Tang},
title = {Intelligent Portfolio Management Based on Deep Deterministic Policy Gradient},
journal = {Journal of Business and Economic Research},
year = {2026},
volume = {2},
number = {2},
doi = {10.63887/jber.2026.2.2.30}
}
EUVIMED