Social Media Development and Multi-Modal Input for Stock Market Prediction: A Review

In the realm of behavioral finance, the influence of information—particularly that disseminated by media sources—plays a pivotal role in shaping stock market prices. This study categorizes social media into three distinct evolutionary phases, each characterized by its media agency and mode of interaction: authoritative media, grassroots media, and digital interactive media. As we transition into the digital economy era, digital interactive media has emerged as the predominant medium for the dissemination of investment-related information in China. Each media type exhibits unique attributes and varying degrees of impact on the stock market. Given its voluminous information output and dual-subject nature, there is a pressing need for digital interactive media to refine its research approaches and methodologies through a systematic review of existing literature. This paper conducts an exhaustive review of 53 seminal journal articles that focus on stock price forecasting, encompassing all three major phases of media development. It critically evaluates the current body of research, highlighting the focal points and limitations from three perspectives: the evolution of stock market media, the processing of information on media platforms, and the development of multi-source heterogeneous models. Looking forward, the paper advocates for the application of NLP techniques to quantify sentiment and the utilization of multi-source heterogeneous large-scale models.