投资者情绪预测已实现波动率——基于Scaled (Bayesian)-PCA模型

Investor Sentiment Forecasting Realised Volatility—Based on the Scaled (Bayesian)-PCA Model

摘要: 本研究通过将贝叶斯统计方法融入投资者情绪的测量中，来提高对中国股市股票价格波动的预测准确性。基于贝叶斯框架的优势，通过整合多来源信息并考虑其相互之间的潜在联系，从而更准确地捕捉和量化投资者情绪的变化。相比于传统的情绪测量技术，贝叶斯方法在样本内外测试中展现了显著的预测性能提升，不仅优于五个主要的综合情绪指标。研究结果表明贝叶斯方法在理解和应对市场情绪波动中的有效性，为中国股市的波动率研究提供了一种新的角度。

Abstract: This study improves the accuracy of predicting stock price volatility in the Chinese stock market by incorporating Bayesian statistical methods into the measurement of investor sentiment. Based on the strengths of the Bayesian framework, changes in investor sentiment are more accurately captured and quantified by integrating information from multiple sources and considering their potential connections with each other. Compared to traditional sentiment measurement techniques, the Bayesian approach demonstrates significant predictive performance enhancement in in-sample and out-of-sample tests, not only outperforming the five main composite sentiment indicators, but also showing higher robustness and predictive power when using a single sentiment proxy variable for prediction. The findings demonstrate the effectiveness of the Bayesian approach in understanding and responding to market sentiment volatility and provide a new perspective for volatility research in the Chinese stock market.