Construction and Empirical Study of a Quantitative Sector Rotation Strategy for A-Shares Based on BERT Sentiment Analysis

DOI: https://doi.org/10.62517/jse.202511614

Author(s)

Lihan Zheng​

Affiliation(s)

Wuhan Business University, Wuhan, China

Abstract

This paper tackles the drawbacks of conventional quantitative investment strategies in the A-share market by making use of the fast growth of financial text data from social media, putting forward an industry rotation quantitative strategy based on BERT sentiment analysis. The study constructs a three-dimensional time-series dataset that combines social media text and financial market data, using a fine-tuned BERT model for sentiment analysis to pull out dynamic industry sentiment indicators. Based on this, a dual-factor dynamic rotation rule that includes macroeconomic cycles is devised, and the strategy's effectiveness is verified through phased backtesting. Empirical findings show that this strategy constantly produces stable excess returns in different market situations, giving new perspectives for investors to optimize portfolios and financial institutions to develop quantitative products. The research not only broadens the scope of quantitative financial market studies but also offers empirical backing for the application of AI technology in the financial field.

Keywords

Chinese A-Share Stock Market; Quantitative Investment Tactics; BERT Sentiment Assessment; Industry Sector Rotation; Social Media Textual Data

References

[1] Wang Ying, Ku Tingting, Xu Shuping, et al. Analysis of Emotional Components in Awe: Text Mining Based on Social Networks [J]. Psychological Technology and Applications, 2020, 8(04): 235-242. [2] Wang Jun, Li Qing. Research on the Impact of Digital Interactive Media on the Stock Market from a Big Data Perspective [M]. Southwest University of Finance and Economics Press: November 2020: p. 219. [3] Xu Xuechen, Tian Kan. A Novel Method for Stock Index Forecasting Based on Sentiment Analysis of Financial Texts [J]. Journal of Quantitative Economics and Technical Economics, 2021, 38(12): 124-145. [4] Xu, T. (2024). A Study on Quantitative Portfolio Strategies Based on the Chinese and U.S. Stock Markets [D]. Zhejiang University of Science and Technology. [5] Ji, Yuwen; Chen, Zhe. Sentiment Analysis and Applications of Financial Texts Based on BERT. Software Engineering, 2023, 26(11): 33-38. [6] Chen Lingcheng. Research and Application of Deep Learning-Based Sentiment Analysis Methods for Financial Texts [D]. Donghua University of Technology, 2022. [7] Markos G C ,Dimitrios G ,Konstantinos K .Deep Learning for Stock Market Prediction Using Sentiment and Technical Analysis[J].SN Computer Science,2024,5(5). [8] Ching P S ,TienPing T ,Yong H C , et al.A Review on Sentiment Analysis in Reinforcement Learning Model for Stock Market Analysis[J].International Journal of Asian Language Processing,2022,32(04). [9] Li X ,Chen L ,Chen B , et al.BERT-BiLSTM-Attention model for sentiment analysis on Chinese stock reviews[J].Applied Mathematics and Nonlinear Sciences,2024,9(1). [10] Bollen J ,Mao H ,Zeng X .Twitter mood predicts the stock market[J].Journal of Computational Science,2011,2(1):1-8. [11] Araci D .FinBERT: Financial Sentiment Analysis with Pre-trained Language Models.[J].CoRR,2019,abs/1908.10063