The price prediction of virtual currency base on improved support vector regression

Abstract

With the emergence of a large number of virtual currencies represented by Bitcoin and Ethereum and the continuous rise of their prices, a large number of investors have been attracted to invest in them. The attention and research based on virtual currencies have aroused the wide concern of the social public. The price prediction of the virtual currencies is one of the research hotspots, obtaining and analyzing the historical data of virtual currency price for the prediction of future price. We find that some commonly used machine learning algorithms have a great deviation in the price prediction of virtual currency. To solve this problem, we propose a novel support vector regression (SVR) based on data segmentation, which can significantly improve the accuracy and effectiveness of price prediction of virtual currency. The experimental results show that our algorithm has obvious advantages over the traditional SVR algorithm and other eight classical machine learning algorithms.