Time Series Classification by Class-Specific Mahalanobis Distance Measures

TL;DR

This paper compares class-specific Mahalanobis distance measures with DTW for time series classification, highlighting speed advantages and proposing methods to improve Mahalanobis measures' effectiveness.

Contribution

It introduces class-specific Mahalanobis distance measures with strategies like covariance shrinking and diagonal approximation, and benchmarks them against DTW and LMNN.

Findings

DTW outperforms Mahalanobis measures in accuracy

Mahalanobis measures are significantly faster

Class-specific Mahalanobis measures with covariance shrinking improve results

Abstract

To classify time series by nearest neighbors, we need to specify or learn one or several distance measures. We consider variations of the Mahalanobis distance measures which rely on the inverse covariance matrix of the data. Unfortunately --- for time series data --- the covariance matrix has often low rank. To alleviate this problem we can either use a pseudoinverse, covariance shrinking or limit the matrix to its diagonal. We review these alternatives and benchmark them against competitive methods such as the related Large Margin Nearest Neighbor Classification (LMNN) and the Dynamic Time Warping (DTW) distance. As we expected, we find that the DTW is superior, but the Mahalanobis distance measures are one to two orders of magnitude faster. To get best results with Mahalanobis distance measures, we recommend learning one distance measure per class using either covariance shrinking or the diagonal approach.