Forecasting Bitcoin Prices: ARIMA and Seasonal Decomposition Approach

Abstract

The research is done to forecast the Bitcoin prices, the study considered Bitcoin prices. The data has been collected in an hourly basis from 1st July 2017 to 1st July 2018. The data is of high frequency. The forecast has been made for a period of 12 months from 31st July 2018 to 31st July 2019. The forecasted data has two different periods, one to calculate the error and the other to have an idea of the future price changes. The study uses Autoregressive Integrated Moving Average(ARIMA) and Seasonal Decomposition method for forecasting the Bitcoin prices. This article also helps the investors to have an idea about the future prices of Bitcoin. The forecasted prices also act as a factor for investor decision making, based on the forecast further choices can be made based on the Bitcoin price changes and invest accordingly. The results of this paper have has sown that, between the two models used, ARIMA has resulted to be the best model compared to the seasonal decomposition model as, the Mean Absolute Error, Mean Absolute Percentage Error and Root Mean Square Error is lower in ARIMA forecasting model, which satisfies the objective of the paper and the best model is selected.