Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Quantifying Qualitative Insights: Leveraging LLMs to Market Predict

Hoyoung Lee, Youngsoo Choi, Yuhee Kwon

TL;DR

This study addresses challenges in Large Language Models by leveraging daily reports from securities firms to create high-quality contextual information and demonstrates that LLMs outperform time-series models in market forecasting.

Abstract

Recent advancements in Large Language Models (LLMs) have the potential to transform financial analytics by integrating numerical and textual data. However, challenges such as insufficient context when fusing multimodal information and the difficulty in measuring the utility of qualitative outputs, which LLMs generate as text, have limited their effectiveness in tasks such as financial forecasting. This study addresses these challenges by leveraging daily reports from securities firms to create high-quality contextual information. The reports are segmented into text-based key factors and combined with numerical data, such as price information, to form context sets. By dynamically updating few-shot examples based on the query time, the sets incorporate the latest information, forming a highly relevant set closely aligned with the query point. Additionally, a crafted prompt is designed to assign scores to the key factors, converting qualitative insights into quantitative results. The derived scores undergo a scaling process, transforming them into real-world values that are used for prediction. Our experiments demonstrate that LLMs outperform time-series models in market forecasting, though challenges such as imperfect reproducibility and limited explainability remain.

Quantifying Qualitative Insights: Leveraging LLMs to Market Predict

TL;DR

This study addresses challenges in Large Language Models by leveraging daily reports from securities firms to create high-quality contextual information and demonstrates that LLMs outperform time-series models in market forecasting.

Abstract

Recent advancements in Large Language Models (LLMs) have the potential to transform financial analytics by integrating numerical and textual data. However, challenges such as insufficient context when fusing multimodal information and the difficulty in measuring the utility of qualitative outputs, which LLMs generate as text, have limited their effectiveness in tasks such as financial forecasting. This study addresses these challenges by leveraging daily reports from securities firms to create high-quality contextual information. The reports are segmented into text-based key factors and combined with numerical data, such as price information, to form context sets. By dynamically updating few-shot examples based on the query time, the sets incorporate the latest information, forming a highly relevant set closely aligned with the query point. Additionally, a crafted prompt is designed to assign scores to the key factors, converting qualitative insights into quantitative results. The derived scores undergo a scaling process, transforming them into real-world values that are used for prediction. Our experiments demonstrate that LLMs outperform time-series models in market forecasting, though challenges such as imperfect reproducibility and limited explainability remain.

Paper Structure

This paper contains 29 sections, 10 equations, 4 figures, 2 tables.

Table of Contents

  1. INTRODUCTION
  2. RELATED WORK
  3. Traditional Methods
  4. LLMs in Finance
  5. Time-series forecasting with LLMs
  6. Prompt Engineering
  7. METHODOLOGY
  8. Data Collection
  9. Multiple Trials Generation
  10. Factor Generation
  11. Autoregressive Moving Shot
  12. Factor Scoring Prompt
  13. Value-to-Score Scaling and Re-Scaling
  14. EXPERIMENTS
  15. Experimental Period
  16. ...and 14 more sections

Figures (4)

  • Figure 1: The methodology involves three steps: generating key factors with a domain-specific model, combining them with price data to create autoregressive moving shots, and using factor scoring prompt for prediction. This process converts qualitative insights into quantitative results.
  • Figure 2: Factor Scoring Prompt is presented with 5-Shot.
  • Figure 3: Heatmap of pairwise correlation coefficients between trials, showing strong consistency with coefficients over 0.85
  • Figure 4: LLM-generated text with Factor Scoring Prompt, along with rationale for transparency.