Stock Trend Prediction using Financial Market News and BERT

Feng Wei, Uyen Nguyen


Stock market trend prediction is an attractive research topic since successful predictions of the market’s future movement could result in significant profits. Recent advances in language representation such as Generative Pre-trained Transformer (GPT) and Bidirectional Encoder Representations from Transformers (BERT) models have shown success in incorporating a pre-trained transformer language model and fine-tuning operations to improve downstream natural language processing (NLP) systems. In this paper, we apply the popular BERT model to leverage financial market news to predict stock price movements. Experimental results show that our proposed methods are simple but very effective, which can significantly improve the stock prediction accuracy on a standard financial database over the baseline system and existing work.


Paper Citation