Core Concepts
Castor introduces a novel shapelet-based time series classification algorithm that outperforms state-of-the-art methods in accuracy and efficiency.
Abstract
Shapelets are discriminative subsequences used for time series classification.
Castor organizes shapelets into groups to compete over temporal contexts.
Features include minimal distance, maximal distance, and occurrence frequency.
Computational complexity is O(log2(m) · m · l).
Castor significantly outperforms Rocket, MultiRocket, Hydra, DST, DrCif, MrSeql, UST, and z-time in accuracy.
Stats
Castorは、HydraとRocketに比べて予測精度が高いことが示されました。
Castorの計算複雑性はO(log2(m) · m · l)です。