Using an Economically Justified Trend for the Stationarity of Time Series in ARMA Models

The ARMA models are used in econometric studies to predict the behavior of a time series. In case of non-stationarity of the initial data ARIMA models get time series to stationarity by differentiation. The problem is applying differentiation provide the loss of essential information. The paper is trying to prove that ARMA model based on the differences between non-stationarity initial data and trend line can provide the same with classic ARIMA approach level of prediction force. For this purpose, the comparison of the quality indicators of the model constructed according to the ARIMA model based on the initial data and the ARMA model based on trend line was carried out. The cryptocurrency market has been chosen as the sphere of research. It was found that the two approaches give approximately the same prediction error and variations from the initial data.