High-dimensional multi-period portfolio allocation using deep reinforcement learning
This paper proposes a deep reinforcement learning framework combining CNN and WaveNet for high-dimensional multi-period portfolio allocation and risk management.
What it examines
Researchers propose a novel multi-period portfolio optimization strategy using deep reinforcement learning. The method employs CNNs to capture dynamic price patterns and WaveNet to model asset dependencies, integrated within a Markov decision process framework with a multi-period Bellman equation. It addresses traditional limitations and adapts to high-dimensional market data effectively.
What it concludes
The study shows that using deep reinforcement learning to choose investments over several periods can make better decisions in complex financial markets by handling risks and costs. It may be used for dynamic asset allocation and algorithmic trading. Future research could refine the model and test it in different markets.
Evidence objects
New portfolio management strategy combines deep reinforcement learning with convolutional neural networks and WaveNet to capture dynamic asset patterns and relationships, delivering superior returns despite transaction costs and variable risks.
key_findings bullet 1 · key_findings · validation V0
Combining feature extraction, asset correlation, and penalty constraints in a multi-period framework using the Bellman equation and Markov decision processes, the method outperforms traditional strategies while managing risk despite volatility.
key_findings bullet 2 · key_findings · validation V0
Extensive tests on indices S&P100, DJIA, and S&P/TSX, using portfolio value, Sharpe ratio, and maximum drawdown metrics, validate the strategy, though heavy reliance on hyperparameter tuning creates real-world deployment challenges.
key_findings bullet 3 · key_findings · validation V0
This paper introduces an original framework that fuses $CNN$-based dynamic price sequence extraction with \$WaveNet\$ for crossasset dependency analysis, embedded in a multi-period decision process via $DRL$ and an extended Bellman equation. Its innovative methodology addresses the challenges of high-dimensional portfolio optimization, rendering it a compelling read for finance researchers.
key_findings bullet 4 · key_findings · validation V0
Raw abstract and provenance
- 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 …
Source row: 1030 · abstract type: snippet