2188. How Generative AI is Revolutionizing Stock Market Predictions in Emerging Markets
Invited abstract in session WD-7: Sustainable investments and financial performance , stream Risk Management in Commodities and Financial Markets .
Wednesday, 14:30-16:00Room: Clarendon GR.01
Authors (first author is the speaker)
| 1. | Mabutho Sibanda
|
| Economics & Management Sciences, North West University |
Abstract
Predicting stock market price movements is crucial for investors, financial institutions, and policymakers to make informed decisions, manage risks, and optimize returns. Traditional models rely on historical price data and technical indicators but often struggle with unstructured data like news and sentiment. The emergence of generative AI (Gen AI) models, such as GPT-4 and FinBERT, has introduced a paradigm shift, enabling the integration of vast textual data, including financial news, earnings reports, and social media sentiment, into predictive frameworks. This study explores the potential of these advanced LLMs in enhancing stock performance prediction, particularly in emerging markets like the Johannesburg Stock Exchange (JSE). Utilising 5-year data of the JSE-Top 40 firms from 2018 to 2023, our results show that LLMs significantly outperform traditional models in capturing sentiment-driven market dynamics, with FinBERT and GPT-4 achieving superior accuracy in processing financial text. Hybrid models combining LLM outputs with quantitative techniques yielded the best predictive results, particularly for medium to long-term forecasting. However, LLMs faced challenges in short-term, high-frequency trading scenarios, and their performance was impacted by market anomalies and low liquidity in emerging markets. These findings highlight the transformative potential of Gen AI in financial forecasting while emphasizing the need for hybrid approaches and ethical considerations.
Keywords
- Artificial Intelligence
- Analytics and Data Science
- Finance and Banking
Status: accepted
Back to the list of papers