USING NONLINEAR MACHINE LEARNING ALGORITHMS TO PREDICT THE PRICE OF CRYPTOCURRENCIES

Abstract

The potential growing for profit in virtual currency has made the prediction of cryptocurrency’s price
an appealing research topic. Numerous research has already been conducted to predict future prices
of a specific virtual currency using a machine-learning model. However, very few involved many
cryptocurrency using various machine-learning model in their studies. This study applies a three nonlinear algorithm: Decision Tree Regressor, (DTR) and the K-Nearest Neighbor (KNN) models for
forecasting the three big cryptocurrency prices: Bitcoin, XRP and Ethereum using bivariate time
series method where the cryptocurrency (daily-Closed Price) is the continuous dependent variable and
the Morgan Stanley Capital International (MSCI) All Country World Index (MSCI-ACWI)-(dailyClosed Price) is the predictor variables. The results demonstrate that (DTR) outperforms the KNearest Neighbor (KNN) in terms of Mean Absolute Error (MAE), Mean Squared Error (MSE), Root
Mean Squared Error (RMSE), and R-Squared (