TC-DTW: Accelerating multivariate dynamic time warping through triangle inequality and point clustering

Dynamic time warping (DTW) plays an important role in analytics on series. Despite the large body of research speeding up univariate DTW, method for multivariate DTW has not been improved much last two decades. The most popular algorithm used today is still one developed nineteen years ago. This paper presents a solution that, as far we know, first consistently outperforms classic across dataset sizes, series lengths, data dimensions, temporal window and machines. new solution, named TC-DTW, introduces triangle inequality point clustering into design lower bound calculations DTW. In experiments DTW-based nearest neighbor finding, avoids 98% (60% average) distance yields 25× (7.5× speedups.