A Darwinian Approach via ML to the Analysis of Cryptocurrencies’ Returns

Abstract This study adopts a Darwinian approach leveraging machine learning (ML) to analyze cryptocurrency returns and their interactions with traditional financial markets. Using a daily dataset from 2018 to 2023, the Random Forest model proved particularly effective in identifying key factors influencing cryptocurrency returns, including technology stock indices (NASDAQ), global equity indices (S&P500, Eurostoxx600), commodity prices (gold, crude oil), and market sentiment (Google Trends). The analysis reveals consistent positive relationships between market sentiment and cryptocurrency returns, highlighting the crucial role of public interest in shaping long-term outcomes. Cryptocurrencies emerge as a distinct asset class with specific correlations to traditional markets and investor sentiment. The study provides strategic insights into understanding cryptocurrency behavior and integrating these dynamics into informed portfolio strategies. It emphasizes the importance of monitoring both traditional financial indices and market sentiment for investment decisions across various time horizons. JEL classification numbers: C58, G11, G15. Keywords: Crypto Assets, Bitcoin, Machine Learning, Investor Decisions.