Sentiment Augmented LSTM for Stock Market Trend Prediction

In this paper, we present a real-time stock price prediction system that combines Long Short-Term Memory (LSTM) neural networks with sentiment analysis of social media data. By integrating numerical stock data with public sentiment, our system aims to improve the accuracy of short-term stock price forecasting. The system ingests live stock price data and social media posts related to specific companies, analyzes sentiment using FinBERT, and trains an LSTM model to predict future stock trends. The architecture utilizes Django, Celery, Redis, PostgreSQL, and Docker to ensure modularity, scalability, and robustness. This approach acknowledges the limitations of relying purely on quantitative indicators and emphasizes the increasing influence of behavioral economics in financial markets. As sentiment reflects the collective psychological state of investors, incorporating it alongside traditional indicators represents a more holistic view of stock market dynamics.