The hardware and bandwidth for this mirror is donated by dogado GmbH, the Webhosting and Full Service-Cloud Provider. Check out our Wordpress Tutorial.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]dogado.de.

geocmeans

An R package to perform Spatial Fuzzy C-means.

R-CMD-check Codecov test coverage

The website of the package is available here

Breaking news

Here we are! We are moving from maptools, sp, rgeos, raster and rgdal to sf, terra and tmap. All the functions and the documentation were modified accordingly. If you spot an error or a bug, please open an issue on github.

Installation

The stable version of geocmeans is available on CRAN. You can install it with the command below.

install.packages("geocmeans")

You can install a development version of the geocmeans package using the command below.

remotes::install_github(repo = "JeremyGelb/geocmeans", build_vignettes = TRUE, force = TRUE)

Authors

Jeremy Gelb, Laboratoire d’Équité Environnemental INRS (CANADA), Email: jeremy.gelb@ucs.inrs.ca

Contributors

Philippe Apparicio, Laboratoire d’Équité Environnemental INRS (CANADA), Email: philippe.apparicio@ucs.inrs.ca

About the package

Provides functions to apply Spatial Fuzzy c-means Algorithm, visualize and interpret results. This method is well suited when the user wants to analyze data with a fuzzy clustering algorithm and to account for the spatial dimension of the dataset. In addition, indexes for measuring the spatial consistency and classification quality are proposed. The algorithms were developed first for brain imagery as described in the articles of Cai and al. 2007 and Zaho and al. 2013. Gelb and Apparicio proposed to apply the method to perform a socio-residential and environmental taxonomy in Lyon (France). The methods can be applied to dataframes or to rasters.

Fuzzy classification algorithms

Four Fuzzy classification algorithms are proposed :

Each function return a membership matrix, the data used for the classification (scaled if required) and the centers of the clusters.

For each algorithm, it is possible to calculate a “robust version” and to add a noise group (used to catch outliers). See the parameters robust and noise_cluser in the documentation for more details.

Parameter selections

The algorithms available require different parameters to be fixed by the user. The function selectParameters is a useful tool to compare the results of different combinations of parameters. A multicore version, selectParameters.mc, using a plan from the package future is also available to speed up the calculus.

Classification quality

Many indices of classification quality can be calculated with the function calcqualityIndexes:

Classification consistency

To assess the stability of the obtained clusters, a function for bootstrap validation is proposed: boot_group_validation. The results can be used to verify if the obtained clusters are stable and how much their centres vary.

Reproductibility

Clustering methods like CMeans depend on the initial centers selected. In geocmeans, they are selected randomly, and two runs of the functions can yield different results. To facilitate the reproductibility of the results, the main functions of the package (CMeans, GFCMeans, SFCMeans, SGFCMeans, selectParameters, selectParameters.mc) have a seed parameter. It can be set by the user to ensure that the results of the functions are exactly the same.

Interpretation

Several functions are also available to facilitate the interpretation of the classification:

There is also a shiny app that can be used to go deeper in the result interpretation. It requires the packages shiny, leaflet, bslib, plotly, shinyWidgets, car.

Alt Text

Spatial diagnostic

Several spatial indices can be calculated to have a better spatial understanding of the obtained clusters, like the global or local Moran I calculated on the membership values, or the join-count-test on the most likely group for each observation. ELSA and Fuzzy ELSA statistics can also be calculated to identify areas with high or low multidimensional spatial autocorrelation in the membership values. See functions spConsistency, calcELSA, calcFuzzyELSA and spatialDiag.

We proposed an index to quantify the spatial inconsistency of a classification (Gelb and Apparicio). If in a classification close observations tend to belong to the same group, then the value of the index is close to 0. If the index is close to 1, then the belonging to groups is randomly distributed in space. A value higher than one can happen in the case of negative spatial autocorrelation. The index is described in the vignette adjustinconsistency. The function spatialDiag does a complete spatial diagnostic of the membership matrix resulting from a classification.

Examples

Detailed examples are given in the vignette introduction

vignette("introduction","geocmeans")

Testing

If you would like to install and run the unit tests interactively, include INSTALL_opts = "--install-tests" in the installation code.

remotes::install_github(repo = "JeremyGelb/geocmeans", build_vignettes = TRUE, force = TRUE, INSTALL_opts = "--install-tests")
testthat::test_package("geocmeans", reporter = "stop")

Contribute

To contribute to geocmeans, please follow these guidelines.

Please note that the geocmeans project is released with a Contributor Code of Conduct. By contributing to this project, you agree to abide by its terms.

License

geocmeans version 0.3.4 is licensed under GPL2 License.

These binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.
Health stats visible at Monitor.