Analyzing Cryptocurrency Social Media for Price Forecasting and Scam Detection

Abstract

The exponential growth of cryptocurrencies in the last decade has been accompanied by a surge in illicit activities, such as pump-and-dump schemes, Ponzi schemes, and exit scams. These nefarious activities have proliferated through the manipulation of social media platforms. This work employs a multimodal approach to predict cryptocurrency price movements by analyzing social media data, search interest, tweet spikes, and historical cryptocurrency prices. Using data from social media discussions, we extract lexical and behavioral features that highlight sentiment shifts, influence tactics, and cross-platform correlations relevant to cryptocurrency price fluctuations. Using a Variational Autoencoder, this approach predicts cryptocurrency prices and unveils patterns of pump-and-dump attempts. With real-time deployment capabilities, this system not only aids investors in understanding and predicting price trends but also offers protection against scams that may coincide with unusual price fluctuations. Moreover, the techniques developed extend to detecting manipulation and organized inauthenticity beyond the realm of cryptocurrency, ensuring trust and security in digital systems.