CRYPTOCURRENCY PRICE FORECASTING FOR FUTURES USING ARIMA MODEL

Abstract

The purposes of this study were to predict the cryptocurrency prices using the ARIMA model and to affirm the prediction accuracy of the cryptocurrency prediction in which the secondary data had been collected weekly from February 18th, 2018 to December 12th, 2021 counted as a total of 200 weeks. The data exactly comprised the 7 digital currencies including Bitcoin, Ethereum, Binance Coin, Cardano, XRP, Dogecoin, and Litecoin. Significantly, the study outcome indicated that: 1) ARIMA (1,1,1) was the optimal forecast model for Bitcoin price since it demonstrated the MAPE of 8.175 and the prediction accuracy of 91.825%; 2) ARIMA (2,1,4) was the optimal model for Ethereum price forecast since it demonstrated the MAPE of 11.192 and the prediction accuracy of 88.808%; 3) ARIMA (0,1,1) was the optimal model for Binance Coin forecast with the MAPE of 15.541 and the prediction accuracy of 84.459%; 4) ARIMA (1,1,0) was the optimal model to predict the price of Cardano with the MAPE of 12.647 and the prediction accuracy of 84.459%; 5) ARIMA (0,1,6) was the optimal model for XRP forecast with the MAPE of 11.886 and the prediction accuracy of 88.114%; 6) ARIMA (2,1,3) was the optimal model for Dogecoin price forecast with the MAPE of 21.852 and the prediction accuracy of 78.148%; and 7) The optimal model for Litecoin price was ARIMA (3,1,2) with the MAPE of 11.059 and the prediction accuracy of 88.059%. After all, the Bitcoin model was the most accurate one whereas the ARIMA model for Dogecoin provided the least accurate result.