Prediction of Cryptocurrency Prices with the Momentum Indicators and Machine Learning

Cryptocurrencies have attracted interest from researchers, investors, and the media. Considering their volatility behaviour, researchers are particularly interested in proposing methods for price prediction. In this paper, we combine technical analysis with a machine learning-based regression to create a price prediction model. We use three momentum indicators (Moving Average Convergence/- Divergence, Commodity Channel Index, Relative Strength Index), the eXtreme Gradient Boosting model, and historical daily data to estimate the close price and stock returns of four cryptocurrencies, namely, Bitcoin, Ether, Golem, and FUNToken.The constructed model demonstrates high accuracy and prediction powers in both training and test datasets. The results of the research provide an accurate way for cryptocurrencies price prediction and contribute to the the literature by providing a combination of traditional technical analysis with machine learning algorithms. This study proves that the utilisation of momentum indicators in machine learning-based predictions can be beneficial for providing an accurate price estimate.