SBCA: Cross-Modal BERT-driven Actor-Critic for Multi-Asset Portfolio Optimization

2026-05-02

AI Summary

This study addresses the limitations of traditional portfolio optimization methods, which rely on linear assumptions, struggle to integrate multimodal data such as price series and financial text, and often ignore real-world trading constraints. To overcome these challenges, the authors propose the SBCA framework, which innovatively combines cross-modal BERT with an Actor-Critic deep reinforcement learning architecture. SBCA employs a gating mechanism to adaptively fuse temporal price dynamics and textual sentiment features, while embedding downside risk control and turnover penalties directly into the reward function to enable end-to-end dynamic multi-asset allocation. Evaluated on eleven years of U.S. multi-asset market data, SBCA significantly outperforms equal-weight, buy-and-hold, and market benchmark strategies in terms of cumulative returns, Sharpe ratio, and maximum drawdown, demonstrating robustness to transaction costs.

Abstract

Portfolio optimization is constrained by linear assumptions and insufficient integration of multi-modal information in traditional models. This paper proposes a cross-modal BERT-driven Actor-Critic framework SBCA for multi-asset portfolio optimization to address the deficiencies of existing deep reinforcement learning DRL methods in fusing price data and financial text sentiment, as well as lacking practical trading constraints. The framework adopts a cross-modal gated fusion mechanism to adaptively integrate price time-series features and text semantic features, embeds downside risk and turnover penalty constraints into the reward function, and constructs a complete empirical system for validation. Experiments on 11-year U.S. stock multi-asset datasets show that SBCA outperforms equal weight, buy-and-hold and market benchmark strategies in portfolio value, annual return, Sharpe ratio and maximum drawdown. Ablation studies verify the complementary enhancement of Actor-Critic mechanism and cross-modal fusion module. Cost sensitivity analysis confirms the model's robustness under varying transaction costs. SBCA provides an effective and interpretable end-to-end solution for dynamic quantitative portfolio decision-making.

Problem

• portfolio optimization
• multi-modal fusion
• financial text sentiment
• trading constraints
• deep reinforcement learning

Innovation

• cross-modal fusion
• BERT-driven reinforcement learning
• portfolio optimization
• downside risk constraint
• actor-critic framework