A Method of Preliminary Forecasting of Time Series of Financial Data

The problem of forecasting the time series of stock prices of leading global companies that are characterized by long-term memory is considered. It is assumed that ignoring the presence of such a correlation structure in time series using traditional methods of analysis leads to a much greater error than taking into account long-term memory in its actual absence. It is assumed that the daily fluctuations in prices for financial market instruments are the Hurst process, that is, they have long-term memory, which means such a time series cannot be effectively analyzed using traditional stationary models that completely ignore this fact. Thus, the task is set, using the R/S-analysis method, to determine the presence of long-term memory in the initial time series and to determine its type.