Big Data Time Series Analysis: An Approach with the Size of the Time Series

Data analysis has importance in real-time data. It is collective representation of all the facts of population or variables under interest. As velocity of data generating process (DGP) is third very important identity in Big Data prospective, so this cannot be managed through routine statistical tools of analysis. Big data time series recording is in principle on packet of information in every unit interval of time and increase size of the series. For that reason modelling can be performed after adding the respective information. This gives us challenges to take a decision regarding size of the series for understanding the data generating process. Present paper is proposing an alternative approach of time series analysis in which data generating process is studies based on size of time series. Here, we discuss the convergence of model in term of order. An empirical analysis is carried out on the time series of sectoral indices of National Stock Exchange (India) to show the convergence of the model. This convergence is also advocating the fact that if a series is fitted after certain size, we get a constant model. This supports that true model forecasted rightly the future value and also data is always convergent is nature. Study reveals that the realization of big data time series may be in considered in respect to size of the series, so identified DGP can be best representative intervals for respective study period. Keywords— Time series; Big Data; Velocity of Big Data; ARIMA model; GARCH model.