Sentiment-based Market Trend Analysis Using Social Media Data

Abstract This study investigates the use of social media sentiment to enhance stock market movement prediction. Platforms like Twitter (X), Reddit, and StockTwits contain emotional cues reflecting public opinion on financial events. By collecting, cleaning, and aligning posts with stock price data, the research examines whether sentiment improves short-term forecasting. Two models—VADER (rulebased) and FinBERT (deep learning)—were used for sentiment scoring. Sentiment indices were then combined with technical indicators and analyzed using ARIMA, XGBoost, and LSTM models. Results show that sentimentenhanced models yield higher predictive accuracy than price-only approaches. The study highlights how integrating public mood signals with market data supports better decision-making in finance. Key Words: Stock prediction, sentiment analysis, FinBERT, VADER, XGBoost, LSTM, financial forecasting.