Evaluation of traditional and machine learning-based investment strategies on cryptocurrency portfolio management

With the popularity of cryptocurrencies increasing, investors are becoming more interested in developing investment portfolios based on or including cryptocurrencies. This often involves using traditional econometric or statistical tools used in more established markets. Yet due to the highly volatile nature of cryptocurrency markets, researchers have been working towards optimised algorithms for incorporating cryptocurrencies into portfolios, often based on machine learning algorithms. Comparisons about the performance of machine learning algorithms and econometrics is however lacking. This study evaluates various portfolio development strategies through backtesting on historical data of 134 cryptocurrencies obtained from online exchange markets. It contributes to understanding about cryptocurrency market behaviour, and how to optimise strategies for incorporating cryptocurrencies within investment portfolios.