Scholarlink

<p>Investments made on the Stock Market (SM) account for a sizeable component of any developing nation's economy. Investors' faith in the SM can be boosted by forecasting the stock price fluctuations accurately, thus augmenting the amount of investment. However, existing time-series analysis techniques were less effective owing to the non-linear, non-stationary, and complex nature of stock price movements. Therefore, in this paper, a novel hybrid-trained Self-Supervised Learning-based Rastrigin Single Candidate Optimized Recurrent Neural Network (SSL-RSCO-RNN) model for SM prediction is proposed. The proposed methodology starts by collecting real-time and historical SM data. The gathered data were preprocessed. After that, by utilizing the A* Heuristics Attribute-Based Data Segmentation (A* HABDS) method, the preprocessed real-time data is segmented centered on the polarity. For extracting features of both data, sentimental feature extraction is conducted. After extraction, by using the L1 regularized- Skip Gram Model (L1-SGM), the word strings in the real-time feature set are converted to vectors. Lastly, for training and predicting the future SM, historical data’s sentimental indexes and features are given to the SSL-RSCO-RNN. In an experimental evaluation, the proposed approach is analogized with baseline models and found to be more effectual than the prevailing approaches.</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 Influenced Deep Learning Model for Stock Market Prediction

  • 期刊论文

  • 发表日期：

    2026年1月29日

查看全文
  • R. Gnanavel,
  • J. M. Gnanasekar

摘要

Investments made on the Stock Market (SM) account for a sizeable component of any developing nation's economy. Investors' faith in the SM can be boosted by forecasting the stock price fluctuations accurately, thus augmenting the amount of investment. However, existing time-series analysis techniques were less effective owing to the non-linear, non-stationary, and complex nature of stock price movements. Therefore, in this paper, a novel hybrid-trained Self-Supervised Learning-based Rastrigin Single Candidate Optimized Recurrent Neural Network (SSL-RSCO-RNN) model for SM prediction is proposed. The proposed methodology starts by collecting real-time and historical SM data. The gathered data were preprocessed. After that, by utilizing the A* Heuristics Attribute-Based Data Segmentation (A* HABDS) method, the preprocessed real-time data is segmented centered on the polarity. For extracting features of both data, sentimental feature extraction is conducted. After extraction, by using the L1 regularized- Skip Gram Model (L1-SGM), the word strings in the real-time feature set are converted to vectors. Lastly, for training and predicting the future SM, historical data’s sentimental indexes and features are given to the SSL-RSCO-RNN. In an experimental evaluation, the proposed approach is analogized with baseline models and found to be more effectual than the prevailing approaches.