High-dimensional multi-period portfolio allocation using deep reinforcement learning
High-dimensional multi-period portfolio allocation using deep reinforcement learning
This paper proposes a novel investment strategy based on deep reinforcement learning (DRL) for long-term portfolio allocation in the presence of transaction costs and risk aversion. We design an advanced portfolio policy framework to model the dynamics of asset prices using convolutional neural networks (CNN), capture group-wise asset dependence using novel WaveNet methods, and solve the optimal asset allocation problem using DRL. These methods are embedded within a multi-period Bellman equation framework. Portfolio performance is tested empirically over different holding periods, risk aversion levels, transaction cost rates, and financial indices. Importantly, the results demonstrate the effectiveness and superiority of the proposed long-term portfolio allocation strategy compared to several competitors based on machine learning methods and under traditional optimization techniques.
Deep reinforcement learning, High-dimensional portfolios, Multi-period portfolio selection, Portfolio constraints, Risk aversion
Jiang, Yifu
cff1ccf8-1299-45de-95ec-f449f30fa0b8
Olmo, Jose
706f68c8-f991-4959-8245-6657a591056e
Atwi, Majed
a713c2fd-6b12-412d-9065-8a72ae788ad7
2 March 2025
Jiang, Yifu
cff1ccf8-1299-45de-95ec-f449f30fa0b8
Olmo, Jose
706f68c8-f991-4959-8245-6657a591056e
Atwi, Majed
a713c2fd-6b12-412d-9065-8a72ae788ad7
Jiang, Yifu, Olmo, Jose and Atwi, Majed
(2025)
High-dimensional multi-period portfolio allocation using deep reinforcement learning.
International Review of Economics and Finance, 98, [103996].
(doi:10.1016/j.iref.2025.103996).
Abstract
This paper proposes a novel investment strategy based on deep reinforcement learning (DRL) for long-term portfolio allocation in the presence of transaction costs and risk aversion. We design an advanced portfolio policy framework to model the dynamics of asset prices using convolutional neural networks (CNN), capture group-wise asset dependence using novel WaveNet methods, and solve the optimal asset allocation problem using DRL. These methods are embedded within a multi-period Bellman equation framework. Portfolio performance is tested empirically over different holding periods, risk aversion levels, transaction cost rates, and financial indices. Importantly, the results demonstrate the effectiveness and superiority of the proposed long-term portfolio allocation strategy compared to several competitors based on machine learning methods and under traditional optimization techniques.
Text
Final manuscript Accepted - Preprint
- Accepted Manuscript
Text
1-s2.0-S1059056025001595-main
- Version of Record
More information
Accepted/In Press date: 26 February 2025
e-pub ahead of print date: 27 February 2025
Published date: 2 March 2025
Keywords:
Deep reinforcement learning, High-dimensional portfolios, Multi-period portfolio selection, Portfolio constraints, Risk aversion
Identifiers
Local EPrints ID: 502201
URI: http://eprints.soton.ac.uk/id/eprint/502201
ISSN: 1059-0560
PURE UUID: e50fb176-a0c5-48b2-b267-1777e3740192
Catalogue record
Date deposited: 18 Jun 2025 16:36
Last modified: 22 Aug 2025 02:08
Export record
Altmetrics
Contributors
Author:
Yifu Jiang
Author:
Majed Atwi
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics
