AI technology for developing Bitcoin investment strategies based on altcoin trends

Abstract

The objective of this study is to analyse the correlation between Bitcoin and altcoins in the post-covid world and take advantage of this possible relationship to design investment strategies on Bitcoin based on the evolution of altcoins using Artificial Intelligence (AI) models.The sample of daily observations covers from January 2020 to February 2023, and the regressions performed between altcoins and Bitcoin are positive and 99 % significant, except for Dogecoin, which has a correlation with Bitcoin.If we add a lag, the estimated parameters are still 95 % significant, except for Dogecoin, so we can assume that the return of altcoins anticipates the evolution of Bitcoin.We train an artificial intelligence model in which the predictors are the observed daily return in altcoins and the target to predict is next day trend of Bitcoin (up or down). We use decision tree algorithms (J48), random forest and naive bayes, but in a retrospective cross-sectional validation with 10 sample partitions we obtain a poor predictive capacity of only a 51 % success rate in the best of cases.Therefore, despite the evident correlation between predictors and the objective variable, we should not implement this investment strategy.

Keywords

• Artificial intelligence
• Cryptocurrencies
• Bitcoin
• Altcoins
• Retrospective validation