Modeling News Interactions and Influence for Financial Market Prediction

FININ, a market prediction model, integrates multi-modal information from market data and news articles, outperforming advanced models in daily Sharpe ratio by 0.429 and 0.341. It reveals insights into delayed market pricing, long memory effect, and limitations of financial sentiment analysis. The model uses a data fusion encoder to process multi-modal inputs, enhancing its ability to understand complex market dynamics. FININ demonstrates better ability to build relationships between news and market movements, improving understanding of market response to news.