This project was developed for the course of Financial Econometrics. The goal of this project is to analyze the volatility of the BTC-USD series using models from the GARCH family using a machine learning approach. Our work is structured in three parts:

1. Data Analysis: in this part we concentrate on data retrieving and on calculating relevant statics of the series of BTC-USD to understand if a GARCH model could be applied and could give desirable results.
2. In-Sample analysis: in this section of the project we have applied several GARCH models (GARCH, EGARCH, GJR-GARCH) and using a restricted dataset we have compared the performance of these models in term of fitted volatility, comparing the latter with realized volatility obtained using hourly data.
3. Out-of-Sample analysis: this last part of the project aim to assess which model is the best to forecast the BTC-USD return volatility. This analysis has been done using a validation set of one year, on which models were trained, and using a test set of five month on which the model were tested.