Comparative Analysis of LLM-Based Market Prediction and Human Expertise with Sentiment Analysis and Machine Learning Integration

EasyChair Preprint 15587

Abstract

This study conducts a comparative analysis of market prediction accuracy between Large Language Model (LLM)-based systems and human expertise within the financial analysis domain. Leveraging Quantum, an advanced LLM specialized for financial forecasting, we evaluate its predictive performance against human analysts and general-purpose LLMs, including GPT-3, GPT-4, FinGPT, and FinBERT. Employing a dataset of historical financial data, news headlines, and social media sentiment, we systematically assess predictive accuracy, response efficiency, and interpretability across models. The integration of sentiment analysis and machine learning further strengthens prediction reliability. Results reveal that Quantum’s specialized model demonstrates superior accuracy and speed in financial forecasting compared to human predictions and generalized LLMs, particularly in fast-moving, data-rich contexts. Nevertheless, limitations in nuanced contextual understanding and adaptability persist, highlighting the enduring value of human expertise. This research reinforces the potential of LLMs as robust tools for financial decision-making while identifying key areas for refinement to enhance synergy with human analytical insights.

Keyphrases: Artificial Intelligence, Financial Prediction, Large Language Models (LLMs), Market Forecasting, Sentiment Analysis, machine learning

@booklet{EasyChair:15587,
  author    = {Mohamed Abdelsamie and Hua Wang},
  title     = {Comparative Analysis of LLM-Based Market Prediction and Human Expertise with Sentiment Analysis and Machine Learning Integration},
  howpublished = {EasyChair Preprint 15587},
  year      = {EasyChair, 2024}}