Scholarlink

<p>The stock market dynamics are intricately linked to market sentiment. Social media for stock market prediction has been questioned for reliability and accuracy, which depend on the learning methodology and the parameter values. This study aims to determine whether public sentiments and technical indicators provide additional predictive power while considering the range of features. Artificial Neural Network (ANN) and Gated Recurrent Unit-Recurrent Neural Network (GRU-RNN) architecture are introduced for predicting stock prices across the comprehensive dataset comprising market data, technical indicators, and sentiments from X over the period from 2016 to 2021. Using the Nifty50 Index of the Indian stock market, this architecture incorporates a unique approach to feature selection. The results of this study demonstrate significant improvements in evaluation metrics, primarily validation Mean Absolute Percentage Error (MAPE) of 8.379% when the ANN model is directly applied to the dataset. A decrease of 1% was observed with hyperparameter tuning of the model. The MAPE score is further reduced to 5% when GRU-RNN was applied to significant features. In addition, other performance metrics were also significantly improved, indicating the superior performance of our models. The achieved results challenge the semi-strong form of the Efficient Market Hypothesis by unleashing the market inefficiencies. This study also highlights the critical role of feature selection in refining prediction accuracies.</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

Sentiment-Driven Improved Stock Prediction: An Evidence from India

  • 期刊论文

  • 发表日期：

    2026年6月26日

查看全文
  • Nusrat Rouf,
  • Atul Negi,
  • Majid Bashir Malik,
  • Akib Mohi Ud Din Khanday,
  • Mudavath Ravi

摘要

The stock market dynamics are intricately linked to market sentiment. Social media for stock market prediction has been questioned for reliability and accuracy, which depend on the learning methodology and the parameter values. This study aims to determine whether public sentiments and technical indicators provide additional predictive power while considering the range of features. Artificial Neural Network (ANN) and Gated Recurrent Unit-Recurrent Neural Network (GRU-RNN) architecture are introduced for predicting stock prices across the comprehensive dataset comprising market data, technical indicators, and sentiments from X over the period from 2016 to 2021. Using the Nifty50 Index of the Indian stock market, this architecture incorporates a unique approach to feature selection. The results of this study demonstrate significant improvements in evaluation metrics, primarily validation Mean Absolute Percentage Error (MAPE) of 8.379% when the ANN model is directly applied to the dataset. A decrease of 1% was observed with hyperparameter tuning of the model. The MAPE score is further reduced to 5% when GRU-RNN was applied to significant features. In addition, other performance metrics were also significantly improved, indicating the superior performance of our models. The achieved results challenge the semi-strong form of the Efficient Market Hypothesis by unleashing the market inefficiencies. This study also highlights the critical role of feature selection in refining prediction accuracies.