Forecasting Chinese Overnight Stock Index Movement Using Large Language Models with Market Summary

Forecasting financial market movement constitutes a complex and pivotal research area within the realm of Financial Technology (Fintech). In this work, we investigate the ability of large language models to predict Chinese overnight stock index movement, utilizing market summary gleaned from news media sources. We fine-tune various pre-trained models to compare the performance with that of Generative Pre-training Transformer (GPT) models, specifically GPT-3.5 and GPT-4, as provided by OpenAI. The empirical findings underscore that the fine-tuned pre-trained models, characterized by fewer parameters and more straightforward architectures, surpass the esteemed GPT-3.5 and GPT-4 models in predictive metrics of accuracy and f1. All fine-tuned models are publicly available on the huggingface platform ( https://huggingface.co/hw2942 ).