Fully convolutional networks with shapelet features for time series classification

In recent years, time series classification methods based on shapelet features have attracted significant research interest because they are interpretable. Although researchers have studied shapelet features for decades, the time complexity of the shapelet extracting process remains high, and the accuracy rates of their methods are not ideal. This study combines a fully convolutional network with shapelet features to address these problems. First, some discriminative subsequences are effectively selected as shapelet features. The original time series is then transformed into shapelet feature vectors. Finally, a fully con-volutional network classifier is trained for the transformed vectors. Experimental results on various datasets demonstrate that the proposed method can achieve high accuracy and extract shapelet features more effectively.(c) 2022 Elsevier Inc. All rights reserved.

Automated summary

This study combines a fully convolutional network with shapelet features to address the low efficiency and inadequate accuracy of shapelet feature extraction in time series classification. Experimental results demonstrate that the proposed method achieves high accuracy and more effectively extracts shapelet features.