Abstract:Time series classification assigns predefined labels or classes to sequences of data points ordered chronologically, which is a fundamental task for time series analysis. Existing time series classification methods mainly focus on a specific type of time series (i.e., univariate time series or multivariate time series), while failing to support both of them efficiently and effectively. In addition, most of existing multivariate time series classification methods model all variables collectively, resulting in protracted computational times and suboptimal accuracy. In this paper, we introduce TSec, an innovative time series classification framework that exhibits high training efficiency and classification accuracy for both univariate time series and multivariate time series. During online classification, TSec first involves sequence segmentation and de-duplication, and then employs pre-trained models to perform classifications. To opti-mize the classification performance, TSec (i) utilizes correlation analysis to reveal closely interconnected groups of variables within multivariate time series data; (ii) incorporates time series alignment and different sliding windows to generate potential shapelets; (iii) applies PAA and SAX techniques to eliminate duplicates, thereby enhancing the quality of shapelets; (iv) adopts Bi-GRU and GCN-GRU models to effectively capture the characteristics of the two types of time series. Extensive experiments on 112 public univariate time series datasets and 26 public multivariate time series datasets show that TSec can achieve both high efficiency and accuracy compared with the state-of-the-art 19 toolkits.

A Feature Segment Based Time Series Classification Algorithm

Integrating Data-Driven Segmentation, Local Feature Extraction and Fisher Kernel Encoding to Improve Time Series Classification

A novel piecewise linear segmentation for time series

Segment-Based Features for Time Series Classification

Multivariate time series classification based on μσ-DWC feature and tree-structured M-SVM

Discriminative Extraction of Features from Time Series

A deep graph kernel-based time series classification algorithm

TSec: an Efficient and Effective Framework for Time Series Classification

PFC: A Novel Perceptual Features-Based Framework for Time Series Classification.

Feature-based Online Segmentation Algorithm for Streaming Time Series (Short Paper).

Time Series Classification with Random Temporal Features

Joint segmentation and classification of time series using class-specific features

Adaptive Time Series Segmentation Algorithm Based on Trend Turning Points and State Changes

Time Series Classification with Shapelet and Canonical Features

An Overview On Feature-Based Classification Algorithms For Multivariate Time Series

An Interpretable Time Series Classification Approach Based on Feature Clustering.

A Study of Modelling Non-Stationary Time Series Using Support Vector Machines with Fuzzy Segmentation Information

A Shapelet Selection Algorithm for Time Series Classification: New Directions.

A Hybrid Segmentation Method for Multivariate Time Series Based on the Dynamic Factor Model

Highly Comparative Feature-Based Time-Series Classification

Research On Machine Learning Algorithms And Feature Extraction For Time Series