Time2Graph: Revisiting Time Series Modeling with Dynamic Shapelets

TL;DR

Time2Graph introduces a novel approach to time series modeling by extracting time-aware shapelets and capturing their evolution through a graph structure, enhancing interpretability and performance across diverse datasets.

Contribution

The paper proposes a new method combining time-aware shapelet extraction with evolution graph construction for improved time series analysis.

Findings

Outperforms 17 state-of-the-art baselines

Effective on multiple public and real-world datasets

Enhances interpretability of time series models

Abstract

Time series modeling has attracted extensive research efforts; however, achieving both reliable efficiency and interpretability from a unified model still remains a challenging problem. Among the literature, shapelets offer interpretable and explanatory insights in the classification tasks, while most existing works ignore the differing representative power at different time slices, as well as (more importantly) the evolution pattern of shapelets. In this paper, we propose to extract time-aware shapelets by designing a two-level timing factor. Moreover, we define and construct the shapelet evolution graph, which captures how shapelets evolve over time and can be incorporated into the time series embeddings by graph embedding algorithms. To validate whether the representations obtained in this way can be applied effectively in various scenarios, we conduct experiments based on three public time series datasets, and two real-world datasets from different domains. Experimental results clearly show the improvements achieved by our approach compared with 17 state-of-the-art baselines.