PREDICTING THE PRICE OF CRYPTOCURRENCY USING SUPPORT VECTOR REGRESSION METHODS

Abstract

The rising profit potential in virtual currency has made forecasting the prices of crypto currency a fascinating subject of study. Numerous studies have already been conducted to predict future prices of a specific virtual currency using a machine-learning model. However, very few have focused on using different kernels of a "Support Vector Regression" (SVR) model. This study applies the Linear, Polynomial and "Radial Basis Function"(RBF) kernels to predict the prices of the three major crypto currencies, Bitcoin, XRP and Ethereum, using a bivariate time series method employing the cryptocurrency (daily-Closed Price) as the continuous dependent variable and the "Morgan Stanley Capital International" (MSCI) World Index (MSCI-WI) and the (daily-Closed Price) as the predictor variable. The results demonstrated that 'RBF' outperforms most other kernel methods in predicting cryptocurrency prices in terms of "Mean Absolute Error"(MAE), "Mean Squared Error" (MSE), "Root Mean Squared Error" (RMSE) and R-squared (í µí±).