Change Direction Filtering Method of Nonstationary Time Series Using an Improved Sine Wave Indicator

Abstract

This paper presents a method for filtering the change direction of nonstationary time series using the improvement type sine wave indicator. The sine wave indicator is a method for converting a time series into a sine curve and calculating where the current value is on the sine curve. However, this method has a problem that nonstationary time series in which a plurality of periodic functions and trend line are mixed cannot be analyzed. Therefore, after removing the high frequency noise from the analysis time series, it is decomposed into the intrinsic mode function for each frequency according to the empirical mode decomposition, and the sine wave indicator is applied to the decomposed time series. By evaluating the sine curve created by the sine wave indicator for the section where the influence of the trend component is small, the current change direction of the nonstationary time series could be filtered.