Utilization of tree-based methods for cryptocurrency price prediction

This thesis evaluates tree-based models in terms of their performance in predicting Bitcoin price movements, and also examines whether the cryptocurrency market, represented by Bitcoin, follows the principles established by the efficient markets hypothesis. The target variable indicates whether a long or short position should be opened during trading and closed the following day. The independent variables used in the construction of the models are technical indicators, macroeconomic factors and other variables. The basic model is the ARIMA model, implemented in the R environment. Tree-based models are developed in Python using the scikitlearn library. All models are shown to have low accuracy and are unable to significantly outperform random predictions.