Time2Graph: Revisiting Time Series Modeling with Dynamic Shapelets

Authors: Ziqiang Cheng, Yang Yang, Wei Wang, Wenjie Hu, Yueting Zhuang, Guojie Song3617-3624

AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental 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 16 state-of-the-art baselines.
Researcher Affiliation Collaboration College of Computer Science and Technology, Zhejiang University State Grid Huzhou Power Supply Co. Ltd, China Key Laboratory of Machine Perception, Ministry of Education, Peking University
Pseudocode Yes Algorithm 1 Shapelet Evolution Graph Construction; Algorithm 2 Time Series Embedding Framework
Open Source Code Yes the source codes, along with the implementation details, parameter settings and documentations, can be found on the project homepage: https://petecheng.github.io/Time2Graph.
Open Datasets Yes We use three public datasets, Earthquakes (EQS), Worms Two Class (WTC) and Strawberry (STB) from the UCR Time Series Archive (Dau et al. 2018)
Dataset Splits Yes We choose XGBoost (Chen and Guestrin 2016) as the outer classifier, and use 5-fold nested cross-validation to conduct fine-tuning on hyper-parameters.
Hardware Specification No The paper does not specify the hardware (e.g., CPU, GPU models, memory) used for running the experiments.
Software Dependencies No The paper mentions using "XGBoost" but does not specify version numbers for XGBoost or any other software dependencies.
Experiment Setup Yes We examine the sensitivities of three important hyperparameters: number of selected shapelets K, graph embedding size B and segment length l.