Random Dilated Shapelet Transform: A New Approach for Time Series Shapelets
This improves time series classification for domains requiring interpretability, though it appears incremental as it builds on existing shapelet methods.
The paper tackles the problem of shapelet-based time series classification being outperformed by recent state-of-the-art methods by introducing a new formulation with dilation and a new shapelet feature, achieving comparable accuracy to state-of-the-art approaches on 112 datasets while maintaining scalability and interpretability.
Shapelet-based algorithms are widely used for time series classification because of their ease of interpretation, but they are currently outperformed by recent state-of-the-art approaches. We present a new formulation of time series shapelets including the notion of dilation, and we introduce a new shapelet feature to enhance their discriminative power for classification. Experiments performed on 112 datasets show that our method improves on the state-of-the-art shapelet algorithm, and achieves comparable accuracy to recent state-of-the-art approaches, without sacrificing neither scalability, nor interpretability.