dynamic time warping step pattern

Asked Sep 18 '23 at 09:01

Active Sep 18 '23 at 09:41

Viewed 22 times

I am measuring the distance between two time series that are of different length. Time series were built by creating a score of windows of years:

t1 = (0,0.5,0.6,..,0.3) of length 72 representing years 1901-1931,1902-1932..

t2 = (0,0.3,0.8,..,0.2) of length 82 representing years 1901-1941,1902-1942..

Series are expected to be very close, but the purpose of using DTW is to prove that using different window length (30,40 years) won't affect the score system.

Should I use an asymetric step pattern when measuring the distance?

edited Sep 18 '23 at 09:41

asked Sep 18 '23 at 09:01

Jinane Jouni

I don't understand your variables t1 and t2. You are writing c(0,0.5,0.6,..,0.3), and why isn't this an increasing sequence? – Sextus Empiricus Sep 18 '23 at 09:31
Sorry the c(..) is what I use in an rscript. No it is not an increasing sequence. each value represents a score value between 0 and 1. so the first value 0 is the score for period "1901-1931", "0.5" for "1902-1932"... – Jinane Jouni Sep 18 '23 at 09:43
It still doesn't make sense. – Sextus Empiricus Sep 18 '23 at 10:30

dynamic time warping step pattern

0 Answers0