DTW (Dynamic Time Warping) requires prior normalization?

Question

I'm trying DTW from mlpy, to check similarity between time series.

Should I normalize the series before processing them with DTW? Or is it somewhat tolerant and I can use the series as they are?

All time series stored in a Pandas Dataframe, each in one column. Size is less than 10k points.

score 13 · Answer 1 · answered Jan 03 '17 at 05:32

13

I am glad you asked ;-)

In 99% of cases, you must z-normalize.

Want to know why? I wrote a tutorial on this, page 46 http://www.cs.unm.edu/~mueen/DTW.pdf

answered Jan 03 '17 at 05:32

eamonn

131
2

Any easy way to do this z-normalizaiton in Python? – KcFnMi Jan 03 '17 at 10:00
And, by the way, nice presentation! – KcFnMi Jan 03 '17 at 10:17
@KcFnMi z-normalization in Python: https://docs.scipy.org/doc/scipy-0.14.0/reference/generated/scipy.stats.mstats.zscore.html – Nikolas Rieble Jan 05 '17 at 08:56
@eamonn Awesome slides, thanks for sharing your experience, I can't up vote the answer enough – Emer Mar 14 '19 at 14:59
@eamonn What if the data is multimodal? – Alex Aug 22 '20 at 22:35

score 5 · Accepted Answer · edited Apr 13 '17 at 12:44

DTW often uses a distance between symbols, e.g. a Manhattan distance $(d(x, y) = {\displaystyle |x-y|} $). Whether symbols are samples or features, they might require amplitude (or at least) normalization. Should they? I wish I could answer such a question in all cases. However, you can find some hints in:

DTW (Dynamic Time Warping) requires prior normalization?

2 Answers2

Linked