Outlier Detection Method for Time Series Based on the Rate of Signal Change

To eliminate the influence of outlier for system identification and improve the effect of data preprocessing in time serials, a simple and fast outlier detection and recovering algorithm was presented in this paper. The proposed algorithm, which included the variation rate of signal calculation, statistical parameters estimation, the position of outlier identification, and signal recovering based on interpolation, combined the “rate of signal variation”and statistic theory.Resultsshow that the proposed algorithm has the benefits of simpleness, effectiveness and low computational cost and can be satisfied for the demand of data preprocessing in time series.