top of page

Bridging Sentiment and Strategy: Insights and Future Directions

The stock market is a dynamic system influenced by countless factors, with financial news sentiment playing a significant role in shaping investor behavior. This project set out to explore how advanced neural networks can leverage news sentiment and historical stock data to predict Nvidia's stock prices. Through the use of LSTMs, GRUs, and CNN-LSTM hybrid models, this study demonstrated the potential of combining qualitative insights from sentiment analysis with quantitative market data. The results provided compelling evidence that these models can effectively capture patterns and relationships in complex, time-sensitive datasets.

Screenshot 2024-12-08 at 9.50.30 PM.png
Screenshot 2024-12-08 at 9.49.50 PM.png

Sentiment analysis revealed a predominance of positive headlines in the dataset, reflecting favorable market sentiment toward Nvidia during the analysis period. By quantifying this sentiment, the models were able to incorporate emotional and psychological market drivers into stock price predictions. The GRU model emerged as the most effective, achieving the lowest error metrics. Its efficiency and simplicity made it an excellent tool for this task, proving that more complex architectures are not always necessary for high accuracy.

The inclusion of hybrid models like CNN-LSTM with attention showcased the value of integrating multiple modeling techniques. While the GRU model excelled in this project, the CNN-LSTM's ability to capture both local and long-term patterns remains a promising avenue for future research. The project's findings suggest that even subtle changes in sentiment, as captured by tools like FinBERT, can influence market movements and provide actionable insights for investors.​

Screenshot 2024-12-08 at 9.53.01 PM.png
Screenshot 2024-12-08 at 9.53.42 PM.png

Beyond the technical results, this project highlights the broader potential of combining AI and finance. By bridging the gap between qualitative data (news sentiment) and quantitative data (stock prices), this approach empowers investors, analysts, and institutions to make more informed decisions. It underscores the importance of innovation in financial modeling and the need for adaptable tools to navigate an ever-changing market landscape.

In conclusion, this project demonstrates the power of neural networks and sentiment analysis in unraveling the complexities of the stock market. The insights generated here pave the way for future applications in predictive analytics, risk management, and trading strategies. By making sense of the relationship between news and market trends, this study offers a glimpse into how AI can transform financial decision-making for the better.

Screenshot 2024-12-08 at 9.54.28 PM.png

Future Work

While this project achieved promising results, there are several directions for future research:

  • Incorporating Real-Time Data: Integrate live news feeds and stock updates to enable real-time stock price predictions.

  • Sector-Specific Analysis: Expand the model to analyze multiple sectors and explore how sentiment impacts different industries.

  • Macro-Economic Features: Include broader economic indicators like GDP, inflation rates, and market indices to enhance model robustness.

  • Explainability Techniques: Employ methods like SHAP values to make model predictions more interpretable for stakeholders.

  • Advanced Attention Mechanisms: Experiment with Transformer-based architectures like BERT or GPT to further refine sentiment analysis and prediction accuracy.

bottom of page