Portfolio constructions in cryptocurrency market: A CVaR-based deep reinforcement learning approach

Tianxiang Cui, Shusheng Ding, Huan Jin, Yongmin Zhang

Research output: Journal PublicationArticlepeer-review

12 Citations (Scopus)

Abstract

Cryptocurrency markets have much larger tail risk than traditional financial markets, and constructing portfolios with such large tail risk assets would be challenging. Therefore, cryptocurrency funds demand new superior risk management models and Conditional Value at Risk (CVaR) is a prevailing risk measure for constructing portfolios in stock markets with large tail risk. Consequently, our paper contributes to the literature by developing a new cryptocurrency portfolio model framework based on the CVaR risk measure and a deep reinforcement learning optimization framework. We use the data from cryptocurrency market starting 2015 to 2021, unfolding that CVaR measure with deep learning outperforms the traditional portfolio construction technique. Compared with traditional economic parameter-based portfolio models, our model free based approach can capture the nonlinear compounding effect of multiple risk shocks by deep reinforcement learning on the risk distribution with economic structural breakdown. It can guide investments in financial markets with high tail risks.

Original languageEnglish
Article number106078
JournalEconomic Modelling
Volume119
DOIs
Publication statusPublished - Feb 2023

Keywords

  • Cryptocurrency market
  • Efficient frontier
  • Portfolio optimization
  • Reinforcement learning

ASJC Scopus subject areas

  • Economics and Econometrics

Fingerprint

Dive into the research topics of 'Portfolio constructions in cryptocurrency market: A CVaR-based deep reinforcement learning approach'. Together they form a unique fingerprint.

Cite this