Linear Time Complexity Time Series Clustering with Symbolic Pattern Forest

Authors: Xiaosheng Li, Jessica Lin, Liang Zhao

IJCAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate the proposed algorithm extensively on all 85 datasets from the well-known UCR time series archive, and compare with the state-of-the-art approaches with statistical analysis.
Researcher Affiliation Academia 1Department of Computer Science, George Mason University, USA 2Department of Information Science and Technology, George Mason University, USA
Pseudocode Yes Algorithm 1 gives the pseudo-code of SPF.
Open Source Code Yes The C++ source code of SPF is available in the supplementary material1. 1http://mason.gmu.edu/~xli22/SPF
Open Datasets Yes To evaluate the proposed algorithm, we run it on all 85 datasets from the UCR time series archive [Chen et al., 2015]. This public archive contains different types of labeled time series from various fields.
Dataset Splits No Each dataset in the archive contains a training set and a testing set. We fuse both sets and use all the data in the experiment.
Hardware Specification Yes A single core of AMD Opteron Processor 6276 (2299 MHz) and 16 GB memory are used.
Software Dependencies No The C++ source code of SPF is available in the supplementary material1. In our implementation, we use Metis [Karypis and Kumar, 1998] to partition the graph. The source code of k-shape and k-means are obtained from the author of [Paparrizos and Gravano, 2015] and the code is in Matlab. (No specific version numbers for C++, Metis, or Matlab are provided.)
Experiment Setup Yes The number of iterations of k-shape and kmeans are set to 100... The ensemble size of SPF is set to 100 in all the experiments. k is set to equal the number of classes of the dataset in use. ...In SPF, γ is set to 4... ω takes the values from wd = {3, 4, 5, 6, 7} and l takes the values from wl = {0.025, 0.05, 0.075, . . . , 1}m... we set the lower bound to 0.25 n/k... The upper bound is set to n 0.25 n/k.