Year: 2025
Type: article
Abstract: News sentiment can reflect investors’ attitudes toward specific news and their potential investment adjustments based on that sentiment. News sentiment strongly influences investor behavior, offering key advantages to financial market participants. This article uses news sentiment analysis as a new tool and tries to build a portfolio based on this sentiment to test how it works in the capital market. For listed companies, stock prices serve as a clear indicator shaped by investor sentiment—often driven by news coverage. This study uses a CNN-LSTM model to forecast stock prices of target firms and employs FinBERT for sentiment analysis of news from major listed companies (2021–2024). Results show the CNN-LSTM model achieves higher prediction accuracy than methods using only market data or FinBERT. Additionally, a daily rebalanced portfolio was built using the Black-Litterman model, integrating predicted returns and market equilibrium. This portfolio delivered substantial annualized returns and a high Sharpe ratio. The framework provides investors with a novel strategy and offers researchers innovative analytical tools for financial time-series and capital markets.
Authors: Weiye Ju, Qiuyao Jiang
Institutions: University of Southampton, Shanghai University
Language: English
Field: Decision Sciences
Domain: Social Sciences
Open Access status: hybrid
OpenAlex ID: W4412184134