A Self-Adaptive Trend Extraction Algorithm: Based on Slope and Relative Displacement
Jinyao Wu, Xuzhai Wang · 2019
The Self-Adaptive Time Series Trend Extraction Algorithm (PLR-SATEA) based on slope and relative displacement extracts trends of time series based on the slope variation amplitude and relative displacement shape of time series data itself. Without setting threshold, PLR-SATEA algorithm can accurately identify portion extreme points, horizontal inflection points and main trends. After one traversal, the portion one-way trend can be removed. Therefore, the algorithm is highly self-adapted, supporting dynamic growth. Its time complexity is O(n). Experimental results show that the PLR-SATEA algorithm has strong universality. In the fitting experiments of 85 UCR data sets collected by Keogh et al., the fitting error, unit time fitting error and compression ratio all performed well.