Forecasting Cryptocurrency Prices Using Ensembles-Based Machine Learning Approach

This paper is devoted to the problems of the short-term forecasting cryptocurrencies time series using machine learning approach. We applied two the most powerful ensembles methods: Random Forests (RF) and Gradient Boosting Machine (GBM). For testing models we used the daily close prices of three the most capitalized coins: Bitcoin (BTC), Ethereum (ETH) and Ripple (XRP), and as a features were selected the past price information and technical indicators (moving average). To check the efficiency of these models we made out-of-sample forecast for three cryptocurrencies by using one step ahead technique. As the accuracy rate for our models we were selected Mean Absolute Percentage Error (MAPE) and Root Mean Square Error (RMSE) metrics. According to comparative analysis of the predictive ability of the RF and GBM both models showed the same order of accuracy for the out-of-sample dataset prediction, although boosting also was somewhat more accurate. Computer experiments have confirmed the feasibility of using the machine learning ensembles approaches considered for the short-term forecasting of cryptocurrencies time series. Built models and their ensembles can be used as the basis algorithms for automated Internet trading systems.