Cryptocurrency price prediction using ARIMA and ARIMAX models

Abstract

This work discusses short and mid term price prediction of Dogecoin, Shiba Inu and Idena cryptocurrencies using ARIMA and ARIMAX models. Data was sampled from various publicly available sources, preprocessed to ensure stationarity and model parameters chosen with grid search and auto arima function. Different prediction intervals were tested and models evaluated using Akaike Infromation Criterion and mean average percent error. For short term predictions ARIMA model had error of about 3.5%, ARIMAX 7 to 19%. Both models proved inadequate for mid term predictions.