Forecasting the Price of Cryptocurrencies and Validating Using Arima

Abstract

With the increase in popularity of cryptocurrencies, it is becoming extremely crucial to predict what the prices of the currencies are going to be in the future. This paper uses a dataset that consists of over 1500 cryptocurrencies with their prices starting from their initiation till May, 2018. A lot of the effort went into getting the data set ready before predicting the future prices of all the cryptocurrencies, i.e., making sure that the cryptocurrencies were stationary time-series. Beginning with learning about the ARIMA model and the conditions to run the model successfully, first validation of the model is done. An average accuracy of 86.424 is observed for 95% of the currencies are observed. After this validation, forecasting is performed on these cryptocurrencies and the percentage change of the price is calculated.