Decision-informed Neural Networks with Large Language Model Integration for Portfolio Optimization

2025-02-02

AI Summary

Existing portfolio management approaches suffer from inefficiency and poor robustness due to misalignment between prediction objectives (e.g., return forecasting) and downstream decision-making goals. Method: We propose a decision-aware end-to-end optimization framework that aligns large language models’ (LLMs) semantic representation capability with portfolio decision objectives. A multi-source attention mechanism jointly models asset interdependencies, temporal dynamics, and macroeconomic variables, while a differentiable portfolio optimization layer enables gradient propagation through the entire pipeline. We theoretically prove that minimizing prediction error does not guarantee optimal decisions, and thus formulate a decision-performance-driven joint training paradigm. Results: Our method achieves significant gains over state-of-the-art deep learning baselines on S&P 100 and DOW 30 datasets. Gradient attribution analysis confirms automatic focus on salient assets, enhancing noise resilience and decision robustness.

Technology Category

• Machine Learning: Optimization
• Natural Language Processing: Learning & Optimization for NLP
• Search and Optimization: Learning to Search

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

This paper addresses the critical disconnect between prediction and decision quality in portfolio optimization by integrating Large Language Models (LLMs) with decision-focused learning. We demonstrate both theoretically and empirically that minimizing the prediction error alone leads to suboptimal portfolio decisions. We aim to exploit the representational power of LLMs for investment decisions. An attention mechanism processes asset relationships, temporal dependencies, and macro variables, which are then directly integrated into a portfolio optimization layer. This enables the model to capture complex market dynamics and align predictions with the decision objectives. Extensive experiments on S&P100 and DOW30 datasets show that our model consistently outperforms state-of-the-art deep learning models. In addition, gradient-based analyses show that our model prioritizes the assets most crucial to decision making, thus mitigating the effects of prediction errors on portfolio performance. These findings underscore the value of integrating decision objectives into predictions for more robust and context-aware portfolio management.

Problem

• Portfolio Management
• Prediction-Decision Inconsistency
• Investment Decision Efficiency

Innovation

• Language Model
• Attention Mechanism
• Portfolio Management