Time Series Analysis by Empirical Mode Decomposition

Empirical Mode Decomposition (EMD) is a data-adaptive method that provides an approach for decomposing a time series. Because of its robustness in the presence of non-linearity and non-stationarity, EMD has received much attention in the past decade and has been used in various fields. EMD decomposes a time series into sub-series called Intrinsic Mode Functions (IMFs) according to the levels of its local oscillation or frequency. This technique automatically extracts oscillations embedded in a time series and efficiently captures non-linear features with respect to amplitude and frequency modulation at local time scale. The aim of this paper is to briefly introduce EMD to researchers interested in time series analyses. Using the R package EMD, we demonstrate promising potentials of EMD for non-stationary time series provided through a real-world data set. This paper is supplemented with accompanying R codes.