Abstract
This study investigates the influence of social media on financial markets, specifically
examining FAANG (Facebook (now Meta), Amazon, Apple, Netflix and Alphabet) and MEME
(GameStop, AMC Entertainment, BlackBerry, Bed Bath & Beyond, and Virgin Galactic
Holdings) stocks across the COVID-19 pandemic (2019-2022). Leveraging advanced Machine
Learning methods such as Decision Trees, XGBoost, along with autocorrelation, Cross-
Sectional Absolute Deviation, Cross-Sectional Standard Deviation, and Runs Tests, it aims to
uncover patterns in the relationship between social media sentiment and stock price
movements.
Results revealed heightened herding behaviour, particularly in MEME stocks during the
pandemic, indicating an increased impact of broader market sentiment on stock prices. The
Machine Learning models demonstrated varied performance, with XGBoost exhibiting
superior metrics in predicting positive and negative instances for both FAANG and MEME
stocks.
This study’s conclusion challenges the absolute efficiency of financial markets during turbulent
periods but emphasise the transient nature of these effects. This study contributes to the
literature by innovatively integrating social media sentiment analysis with Machine Learning
models, enhancing the understanding of how social media-driven herding behaviour affects
market efficiency. By combining regression analysis and advanced Machine Learning
techniques, the research advances relationship testing methodologies, offering new insights
into how online sentiment can influence stock prices and market movements in a measurable
and actionable way. Limitations include computational constraints and the need for
comprehensive datasets. The research sets the stage for future studies to explore social
media's impact on financial markets with more extensive datasets and refined methodologies.