Year: 2024
Type: article
Abstract: In the rapidly evolving field of financial forecasting, the accurate prediction of stock returns remains a significant challenge. This paper aims to leverage Natural Language Processing algorithms (NLP) to develop a predictive model for stock returns. The research utilized return labels derived from stock price return and sentiment data extracted from StockTwits (a financial social media platform) comments from January 2020 to March 2022. A comparative analysis was conducted to assess the performance of traditional statistical models (Logistics Regression Model), supervised models (Random Forest , Gradient Boosting, XGBoost and Naïve Bayes Model), and an ensemble model (Majority Vote Model) in the prediction tasks. The objective was to identify the most effective model, and to provide precise predictions for future stock returns. Our simulations show that (1) Sentiments can work as an effective proxy to predict stock return; (2) “likes” from users to comments is suitable for price prediction; (3) Logistic Regression didn’t work well in prediction, even when used with other techniques; (4) Random Forest Model and Gradient Boosting Model outperform other simpler models, showing promising predictive results; (5) Ensemble model effectively diminishes the influences of potential model overfitting problems. These findings underline the potential of sentiment analytic models as a tool for more accurate financial forecasting.
Source: Theoretical and Natural Science
Authors: Xianyin Zhang, Haoran Feng, Shuyu Li, Yiqiao Yang
Language: English
Field: Decision Sciences
Domain: Social Sciences
Open Access status: hybrid
OpenAlex ID: W4405599710