Bitcoin Data Analysis Using Deep Learning and Statistical Modeling

Abstract

Bitcoin is a very attractive financial asset for traders and speculators worldwide. This paper first describes how traditional analysis methods can be applied to build statistical models (ARIMA and GARCH) to forecast future returns of bitcoin trading activity. Then a recurrent neural network (RNN) is constructed and trained for bitcoin time-series data. The RNN is a deep learning model for time series, which demonstrates ability to acquire more pattern information from the data.