Attracting investors seeking distributed investing possibilities, cryptocurrencies are gradually taking front stage on financial markets. But sentiment analysis is essential for understanding market dynamics with considerable volatility molded by news, social media trends, and investor mood. This paper investigates the relevance of Large Language Models (LLMs), particularly fine-tuned GPT-4, in cryptocurrency sentiment analysis. By fine-tuning GPT-4 using a cryptocurrency news dataset, this paper compares its sentiment classification performance against other models, including FinBERT, BERT, Flan-T5, and Gemma-7B. The results indicate higher accuracy since finely tuned LLMs show better in classifying sentiments as positive, neutral, or negative. These results highlight the need of optimizing to raise sentiment analysis capability. This paper contributes to both academic research and financial applications, offering insights into how LLMs can be leveraged for market trend predictions and risk management strategies. |
*** Title, author list and abstract as submitted during Camera-Ready version delivery. Small changes that may have occurred during processing by Springer may not appear in this window.