sangyx/deep-finance

Curated collection organized across six research domains—stock prediction, portfolio selection, risk management, finance NLP, blockchain, and market making—combining benchmark datasets (StockNet, EarningsCall, FiQA), peer-reviewed papers, and foundational books. Emphasizes multi-modal approaches including graph neural networks for equity correlations, recurrent architectures for time-series forecasting, and NLP techniques for sentiment extraction from earnings calls and financial news. Serves as a reference taxonomy for practitioners integrating deep learning with quantitative finance rather than a monolithic framework or toolkit.

1,480 stars. No commits in the last 6 months.