A Systematic Review on Social Media Sentiment in Stock Price Prediction: Data Sources, Machine Learning Models, and the Evidence on Effectiveness

Abstract

As the rapid spread of information through social media has increased and sentiment-driven market movements have become more frequent, it’s necessary to understand how social media influences stock price prediction. This paper presents a systematic literature review (SLR) of recent studies (2020–2024) focusing on stock price prediction that incorporates social media sentiment. The review follows the PRISMA framework, involving the selection and analysis of 22 peer-reviewed articles based on predefined criteria. The review examines the types of social media platforms used, the sources of stock price data, the machine learning models applied, and the limited evidence available on the effectiveness of incorporating social media sentiment into stock price prediction. The findings reveal that X (formerly Twitter) is the most commonly used social media platform for obtaining data, and historical stock prices were primarily retrieved from Yahoo Finance. Moreover, LSTM is the most utilized model across studies, followed by RNN and other deep learning variants. However, only a single research directly compares prediction performance with and without the sentiment data, revealing a 2.2% improvement. This review summarizes current trends, highlights methodological limitations, and proposes directions for future studies, including the integration of transformer-based models and more intensive evaluation frameworks for sentiment impact.