Reproduce “Stock price prediction using BERT and GAN” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Priyank Sonkiya, Vikas Bajpai, Anukriti Bansal
Researchers from The LNM Institute of Information Technology developed S-GAN, a model combining BERT-based sentiment analysis with Generative Adversarial Networks to predict stock prices. The model integrates market sentiment from financial news alongside technical indicators and historical price data, achieving a 50% lower RMSE compared to previous similar work.
Key insights
- Incorporating market sentiment, derived from financial news, significantly improves stock price prediction accuracy when combined with technical indicators and historical data.
- Generative Adversarial Networks (GANs) can effectively capture complex data distributions in financial time series, especially when augmented with sentiment as a latent input.
- The adversarial training process of the S-GAN, combined with sentiment integration, leads to improved prediction accuracy and early convergence compared to deep learning baselines and GANs without sentiment.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.