Scholarlink

<p>This paper introduces Bidirectional Encoder Representations from Transformers-based Financial Sentiment Index Enhanced (BERTFSIE) models for assessing financial market efficiency. Unlike traditional approaches, BERTFSIE integrates investor sentiment and asset log-returns to forecast stock movements. Performance metrics such as mean absolute error, root mean square error, weighted F1-score, and area under the receiver operating characteristic curve (AUROC) demonstrate that BERTFSIE consistently outperforms autoregressive integrated moving average (ARIMA), long short-term memory (LSTM), and hybrid models–particularly in directional prediction accuracy. Statistical validation using the Bonferroni–Dunn test confirms these results. In alignment with behavioral finance theory, this study highlights how cognitive biases and investor sentiment contribute to market inefficiencies, challenging the semi-strong form of market efficiency. By leveraging sentiment-driven signals, BERTFSIE models reveal the limited extent to which stock prices reflect external information, offering robust statistical evidence for both investors and policymakers. Overall, the findings underscore the promise of advanced computational methods in examining behavioral influences on market dynamics and in generating insights for financial forecasting and investment decision-making.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Is the financial market efficient? An advanced time series analysis

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月28日

查看全文
  • Si-Qi Mao,
  • Lin-Xi Meng,
  • Adrian Barbu,
  • Xu-Feng Niu

摘要

This paper introduces Bidirectional Encoder Representations from Transformers-based Financial Sentiment Index Enhanced (BERTFSIE) models for assessing financial market efficiency. Unlike traditional approaches, BERTFSIE integrates investor sentiment and asset log-returns to forecast stock movements. Performance metrics such as mean absolute error, root mean square error, weighted F1-score, and area under the receiver operating characteristic curve (AUROC) demonstrate that BERTFSIE consistently outperforms autoregressive integrated moving average (ARIMA), long short-term memory (LSTM), and hybrid models–particularly in directional prediction accuracy. Statistical validation using the Bonferroni–Dunn test confirms these results. In alignment with behavioral finance theory, this study highlights how cognitive biases and investor sentiment contribute to market inefficiencies, challenging the semi-strong form of market efficiency. By leveraging sentiment-driven signals, BERTFSIE models reveal the limited extent to which stock prices reflect external information, offering robust statistical evidence for both investors and policymakers. Overall, the findings underscore the promise of advanced computational methods in examining behavioral influences on market dynamics and in generating insights for financial forecasting and investment decision-making.