Skip to content
/ ZITS Public

This study develops a new clustering method for high-dimensional zero-inflated time-series data by proposing a similarity measure based on Thick Pen Transformation. Two real data set considered were step count data and newly confirmed COVID-19 case data.

Notifications You must be signed in to change notification settings

mjkim1001/ZITS

Repository files navigation

ZITS: Zero-Inflated Time-Series Clustering Via Ensemble Thick-Pen Transform

This GitHub repository contains the R code for the paper titled "Zero-Inflated Time-Series Clustering Via Ensemble Thick-Pen Transform"

by Minji Kim, Hee-Seok Oh and Yaeji Lim, Feb 2023.

The repository includes the following files:

  • functions.R ; Main functions for ZITS clustering, including TPT pen shape and clustering algorithm.

  • simulation_models.R ; Data generation for simulation models.

  • simulation_analysis.R ; ZITS clustering and comparison methods for simulation data, which loads "functions.R" and "simulation_models.R".

  • covid_analysis.R ; ZITS clustering and comparison methods for covid-19 data, which loads "functions.R".

  • step_readData.R ; Data load for step count data.

  • step_analysis.R ; ZITS clustering and comparison methods for step count data, which loads "functions.R" and "step_readData.R".

  • eda_plots.R ; Draw figures for the paper using the step dataset.

The COVID-19 data is publicly available and therefore provided in this repository, but the step count data is not publicly available. If you have any requests regarding the dataset, please contact [email protected].

About

This study develops a new clustering method for high-dimensional zero-inflated time-series data by proposing a similarity measure based on Thick Pen Transformation. Two real data set considered were step count data and newly confirmed COVID-19 case data.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages