Volatility analysis in the cryptocurrency market: from GARCH models to a hybrid approach with Machine Learning

Cryptocurrencies represent one of the most significant financial innovations of the 21st century, reshaping the way markets, institutions and individuals conceive of money and value exchange. Over the past decade, this new asset class has challenged the traditional economic paradigms, to emerge as one of the most dynamic sectors of the global finance. Despite this rapid growth, the market is characterizes by extreme volatility and complex price behaviors, often difficult to capture with traditional financial models, raising the critical questions regarding the stability and the forecasting. Addressing this challenge, the thesis conducts an analysis focused on three major cryptocurrencies: Bitcoin, Ethereum and Binance Coin. It opens with an examination of the historical evolution and characteristics of the crypto market, highlighting the differences respect to the traditional financial assets. The study proceeds with an investigation divide in phases. First, a volatility benchmark is established through the estimation of GARCH’s family models. Second, a hybrid methodology is introduced, combining the best GARCH model with a Machine Learning (ML) algorithm. This work aims to advance the understanding of cryptocurrency volatility, exploring how contemporary hybrid techniques can better describe and predict the market’s complex dynamics.