Leveraging Large Language Models for Predicting Stock Option Valuation and Financial Risk Mitigation

The prediction of short-term stock options with near-future expiration dates is a challenging task due to high volatility, limited information, market noise and the risk of time decay. This work focuses on the new approach to the stock options valuation by leveraging Large Language Models (LLMs) through the integration of quantitative (i.e. financial features-lagged prices, moving averages, and volatility indicators) and qualitative data (i.e. news data, including article titles, full textual content, and publication dates). More specifically, our approach fuses sentiment analysis from LLMs applied to financial news from two reputable outlets (i.e. Economic Times and Yahoo Finance India) with quantitative data on stock options, which includes stock option closing price. By conducting experiments on companies from the NIFTY 50 index using ChatGPT-3.5, ChatGPT-4, and LLaMA 3.1, we show that our method achieves superior prediction accuracy compared to other similar approaches. The paper develops a new framework to improve the valuation of short-term stock options using advanced natural language processing behaviors afforded by LLMs to achieve a more holistic capture of market dynamics and sentiment in option pricing.