Performance Evaluation of Machine Learning Algorithms for Bitcoin Price Prediction

Abstract

Bitcoin is one of the most valuable cryptocurrency in the world with the prices as high as 19,783 United States Dollar(USD) in December of 2017. It made Bitcoin a very profitable market for investment but Bitcoin saw many ups and down as well. Today’s Bitcoin price is 3913 USD but that doesn’t mean that the prices always keep falling. The price of Bitcoin vary over time and is governed by various factors, like the market it is being traded in, scarcity, supply and demand. What has made Bitcoin valuable is that it can be used as a currency, we can pay a part or a fraction of Bitcoin to a person in exchange for something and the part is easily verifiable by blockchain. The small number of Bitcoin, roughly 16 million Bitcoins for the entire world has made it scare and above that its high utility makes it more prestigious.Trading of Bitcoin has proved to be very profitable to many people but the risk in trading is huge as the market of Bitcoin is very volatile. To decrease the risks, this project has been carried out to predict the price of Bitcoin using Recurrent Neural Network(RNN), Long Short Term Memory (LSTM) and Linear Regression(LR) to predict the price of Bitcoin. Evaluation of these algorithms is carried out to determine which among the two is better for the prediction of Bitcoin prices. The dataset used contains minute by minute prices of Bitcoin of over 5 years and contains almost 30,00,000 entries. Since, the dataset used is a big data, evaluating the performance of algorithms over a large dataset will give accurate results.