INTEGRATING FINBERT SENTIMENT WITH FINANCIAL STRESS FOR OPTIMIZED PORTFOLIO PERFORMANCE

This paper introduces an advanced investment strategy for the equity markets, combining a proprietary financial stress index with sentiment analysis conducted using FinBERT, a Bidirectional encoder representation from the Transformers model adapted specifically for financial contexts in deep learning. We analyze sentiments on financial news from HuffPost from 2012 to 2024. Integrating these sentiment analyses into our investment strategy via stress indices significantly improves market forecast accuracy and portfolio performance, achieving higher ratios, and reducing maximum drawdowns compared to conventional strategies. This paper highlights the power of combining advanced machine learning models with extensive financial news data to develop a dynamic trading strategy adaptable across various equity markets. The improved performance is consistent across the NASDAQ, the S&P 500, and the MSCI Global Investment index, indicating that the method generalizes for forecasting and portfolio optimization