Past, present, and future of the application of machine learning in cryptocurrency research

Cryptocurrency has captured the interest of financial scholars and become a major research topic in blockchain. In cryptocurrency research, the use of machine learning algorithms is enabled by the presence of many types of data and abundant resources. However, there is currently no comprehensive review on cryptocurrencies using machine learning. Therefore, we collect papers on cryptocurrency-related using machine learning in the web of science database, and summarise these papers according to the algorithm, and draw the following conclusions: (1) The application of machine learning for cryptocurrencies research is increasing year over year; (2) Predicting cryptocurrency price trends and income fluctuations is the most relevant research topic; (3) The machine learning algorithm utilised in cryptocurrency research is not unique, and the practise of combin ing multiple machine learning approaches has emerged; (4) Concerns such as overfitting and interpretability still persist with machine learning methods.. Finally, we suggest future research directions