New Method of 1-D to 2-D Transformation of Time Series

EasyChair Preprint 3959

Abstract

New method of 1-D to 2-D transformation of time series or one-dimensional signals was proposed. The method is based on a binary transformation of the time series. Resulting binary points are considered as interacting pseudo particles. The transformation is carried out using moving window. In each position of moving window, the pseudopotential is calculated and is written in the central point of moving window. Two pseudopotentials were considered: Ising spin-spin potential and Coulomb potential. Both potentials equally well describe the features of time series dynamics. Potential calculations were carried out at different values of the width of the sliding window. The resulting potential curves were grouped together to create 2-D images. This 1-D to 2_D transformation was compared with continues wavelet transform (CWT). The comparison showed that potential transformation gives more detailed information of hidden patterns then CWT.

Keyphrases: Binarization, Coulomb potential, Detrend, Ising potential, clearing time series from outliers, continues wavelet transformation, normalization, time series

@booklet{EasyChair:3959,
  author    = {Oleg Butusov and Vasily Dikusar},
  title     = {New Method of 1-D to 2-D Transformation of Time Series},
  howpublished = {EasyChair Preprint 3959},
  year      = {EasyChair, 2020}}