Cryptocurrency in global dynamics: Analyzing the Crypto Volatility Index and financial markets with machine learning

Accurate estimation of cryptocurrency market volatility is crucial for investors. The Crypto Volatility Index (CVI) was developed to measure the market's expectations for the 30-day implied volatility of Bitcoin and Ethereum to address the growing demand for reliable predictions. This study explores the relationship between the CVI and the volatility of traditional financial markets, including the Gold Volatility Index (GVZ), the Crude Oil Volatility Index (OVX), and the S&P500 Volatility Index (VIX). Three other variables are also analyzed: the USD to EUR exchange rate (USDEUR), the Federal Reserve interest rate (FED), and the NASDAQ index. The aim of the research is explanatory: the input variables and the CVI are observed contemporaneously to catch the complex relation between them. Using Pearson correlation, distance correlation, and mutual information, we demonstrate the presence of non-linear relationships between some variables in the dataset. Explanatory analysis is conducted using machine learning techniques, specifically the Random Forest (RF) algorithm and Gradient Boosting Machines (GBM) to account for these potential non-linear interactions. These methods are better suited than standard linear models for identifying complex relationships. In particular, the RF algorithm reaches a better level of accuracy than GBM and avoids overfitting.