Statistical Inference for the Dynamic Time Warping Distance, with Application to Abnormal Time-Series Detection
This provides a novel method for quantifying statistical significance in DTW distance, addressing a critical need in abnormal time-series detection for applications like healthcare or finance, though it is incremental as it builds on existing inference frameworks.
The paper tackles the problem of statistical inference for the Dynamic Time Warping (DTW) distance by proposing a conditional selective inference framework to derive valid p-values, enabling high-stake decision-making in abnormal time-series detection, with evaluation on synthetic and real-world datasets.
We study statistical inference on the similarity/distance between two time-series under uncertain environment by considering a statistical hypothesis test on the distance obtained from Dynamic Time Warping (DTW) algorithm. The sampling distribution of the DTW distance is too difficult to derive because it is obtained based on the solution of the DTW algorithm, which is complicated. To circumvent this difficulty, we propose to employ the conditional selective inference framework, which enables us to derive a valid inference method on the DTW distance. To our knowledge, this is the first method that can provide a valid p-value to quantify the statistical significance of the DTW distance, which is helpful for high-stake decision making such as abnormal time-series detection problems. We evaluate the performance of the proposed inference method on both synthetic and real-world datasets.