Scholarlink

Predicting stock prices has perpetually presented a formidable challenge, primarily due to the inherent unpredictability that characterizes financial markets. In response to this challenge, our research project introduces an innovative and sophisticated hybrid approach. This approach harmoniously integrates sentiment analysis and Long Short-Term Memory (LSTM) networks, effectively enhancing the precision of stock price predictions. Our research journey encompasses a comprehensive methodology. We diligently address data preprocessing, employ advanced visualization techniques, conduct sentiment analysis through the utilization of the Valence Aware Dictionary for Sentiment Reasoning (VADER), and implement LSTM networks. This holistic approach ensures that every facet of our model is optimized for performance. To acquire the requisite data, we meticulously collected historical news headlines from Reddit WorldNews, a valuable source of real-world sentiment and events, and curated stock data from the Dow Jones Industrial Average (DJIA), a benchmark for market performance. This meticulous data collection process is vital for the accuracy and relevance of our predictions. Our experimental results underscore the robustness of our approach. When utilizing VADER for label verification, our model showcases a high level of precision, affirming the accuracy and reliability of the labels attributed to our dataset. Furthermore, our LSTM model achieves a remarkable R2 score of 97%, a testament to its capacity to forecast stock prices with an exceptional level of precision. This high R2 score signifies that our model effectively explains 97% of the variation in stock prices, reaffirming its predictive prowess.

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

Stock Price Prediction Using Sentiment Analysis and LSTM Networks

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Mohammed El Idrissi,
  • Nacir Chafik,
  • Ridouane Tachicart

摘要

Predicting stock prices has perpetually presented a formidable challenge, primarily due to the inherent unpredictability that characterizes financial markets. In response to this challenge, our research project introduces an innovative and sophisticated hybrid approach. This approach harmoniously integrates sentiment analysis and Long Short-Term Memory (LSTM) networks, effectively enhancing the precision of stock price predictions. Our research journey encompasses a comprehensive methodology. We diligently address data preprocessing, employ advanced visualization techniques, conduct sentiment analysis through the utilization of the Valence Aware Dictionary for Sentiment Reasoning (VADER), and implement LSTM networks. This holistic approach ensures that every facet of our model is optimized for performance. To acquire the requisite data, we meticulously collected historical news headlines from Reddit WorldNews, a valuable source of real-world sentiment and events, and curated stock data from the Dow Jones Industrial Average (DJIA), a benchmark for market performance. This meticulous data collection process is vital for the accuracy and relevance of our predictions. Our experimental results underscore the robustness of our approach. When utilizing VADER for label verification, our model showcases a high level of precision, affirming the accuracy and reliability of the labels attributed to our dataset. Furthermore, our LSTM model achieves a remarkable R2 score of 97%, a testament to its capacity to forecast stock prices with an exceptional level of precision. This high R2 score signifies that our model effectively explains 97% of the variation in stock prices, reaffirming its predictive prowess.