The DTW algorithm computes the stretch of the time axis which optimally maps one given timeseries (query) onto whole or part of another (reference). It yields the remaining cumulative distance after the alignment and the point-by-point correspondence (warping function). DTW is widely used e.g. for classification and clustering tasks in econometrics, chemometrics and general timeseries mining.
Toni Giorgino
Please see documentation for function dtw(), which is the main
entry point to the package.
The R implementation in dtw provides:
arbitrary windowing functions (global constraints), eg. the
Sakoe-Chiba band; see dtwWindowingFunctions()
arbitrary
transition types (also known as step patterns, slope constraints, local
constraints, or DP-recursion rules). This includes dozens of well-known
types; see stepPattern():
all step patterns classified by Rabiner-Juang, Sakoe-Chiba, and Rabiner-Myers;
symmetric and asymmetric;
Rabiner's smoothed variants;
arbitrary, user-defined slope constraints
partial matches: open-begin, open-end, substring matches
proper, pattern-dependent, normalization (exact average distance per step)
the Minimum Variance Matching (MVM) algorithm (Latecki et al.)
Multivariate timeseries can be aligned with arbitrary local distance
definitions, leveraging the proxy::dist() function of package
proxy. DTW itself becomes a distance function with the dist semantics.
In addition to computing alignments, the package provides:
methods for plotting alignments and warping functions in several classic styles (see plot gallery);
graphical representation of step patterns;
functions for applying a warping function, either direct or inverse; and more.
If you use this software, please cite it according to
citation("dtw"). The package home page is at
https://dynamictimewarping.github.io.
Toni Giorgino. Computing and Visualizing Dynamic Time Warping Alignments in R: The dtw Package. Journal of Statistical Software, 31(7), 1-24. tools:::Rd_expr_doi("10.18637/jss.v031.i07")
Tormene, P.; Giorgino, T.; Quaglini, S. & Stefanelli, M. Matching incomplete time series with dynamic time warping: an algorithm and an application to post-stroke rehabilitation. Artif Intell Med, 2009, 45, 11-34 tools:::Rd_expr_doi("10.1016/j.artmed.2008.11.007")
Rabiner, L. R., & Juang, B.-H. (1993). Chapter 4 in Fundamentals of speech recognition. Englewood Cliffs, NJ: Prentice Hall.
dtw() for the main entry point to the package;
dtwWindowingFunctions() for global constraints;
stepPattern() for local constraints;
proxy::dist(), analogue::distance(), vegan::vegdist() to build local
cost matrices for multivariate timeseries and custom distance functions.