科研之友
微信
新浪微博
Facebook
分享链接
摘要
The increment entropy (IncrEn) is a new measure for quantifying the complexity of a time series. There are three critical parameters in the IncrEn calculation: N (length of the time series), m (dimensionality), and q (quantifying precision). However, the question of how to choose the most appropriate combination of IncrEn parameters for short datasets has not been extensively explored. The purpose of this research was to provide guidance on choosing suitable IncrEn parameters for short datasets by exploring the effects of varying the parameter values. We used simulated data, epileptic EEG data and cardiac interbeat (RR) data to investigate the effects of the parameters on the calculated IncrEn values. The results reveal that Mean is sensitive to changes in in, q and N for short datasets (N <= 500). However, IncrEn reaches stability at a data length of N = 1000 with m = 2 and q = 2, and for short datasets (N = 100), it shows better relative consistency with 2 <= m <= 6 and 2 <= q <= 8 We suggest that the value of N should be no less than 100. To enable a clear distinction between different classes based on IncrEn, we recommend that m and q should take values between 2 and 4. With appropriate parameters, IncrEn enables the effective detection of complexity variations in physiological time series, suggesting that IncrEn should be useful for the analysis of physiological time series in clinical applications.
