A Hybrid Sentiment Analysis Model for Enhanced Stock Market Movement Predictions

Nowadays, stock markets shape global economies and influence investment strategies and individuals' financial plans. Predicting stock trends has become increasingly important for all parties to deal with the complexities of these dynamically changing markets. Social media platforms, especially Twitter, have fundamentally changed market analysis by providing realtime data expressing public sentiment toward companies' performance. This paper leverages vast textual data available on Twitter to enhance stock trend prediction by extracting sentiment. To perform sentiment analysis, a hybrid sentiment analysis model is employed by integrating Financial Bidirectional Encoder Representations from Transformers (FinBert), Valence Aware Dictionary and Sentiment Reasoner (VADER), and KMeans clustering technique. The result of this model, combined with historical stock data, is fed as input to a voting ensemble classifier consisting of Logistic Regression (LR), Random Forest (RF), and Extreme Gradient Boosting (XGBoost) to predict stock trends accurately. The proposed model, evaluated on the StockNet dataset, achieves superior performance compared to previous works on the same dataset, with a precision of 89.77% and an MCC of 0.81.