Predicting the Price of Bitcoin Using Machine Learning

Abstract

The purpose of this article is to assess the feasibility of predicting the future U.S. dollar price of Bitcoin.
The pricing data is obtained from the Bitcoin Price Index. It's a matter of accomplished with varied levels of success by using a Bayesian optimised recurrent neural network (RNN) and a Long Short Term Memory (LSTM) network. The RMSE for the LSTM is 8%, and its classification accuracy is 52%. The famous ARIMA model for time series forecasting \sis constructed as a contrast to the deep learning algorithms. As predicted, the non-linear deep learning algorithms outperform the ARIMA prediction which performs badly. Finally, both deep learning models are benchmarked on both a GPU and a CPU \switch the training time on the GPU surpassing the CPU implementation by 67.7%.